Multiphase Flows in Polymer Microfluidic Systems Namwon Kim Louisiana State University and Agricultural and Mechanical College, Nkim1@Tigers.Lsu.Edu

Louisiana State University LSU Digital Commons

LSU Doctoral Dissertations Graduate School

2009 Multiphase flows in polymer microfluidic systems Namwon Kim Louisiana State University and Agricultural and Mechanical College, [email protected]

Follow this and additional works at: https://digitalcommons.lsu.edu/gradschool_dissertations Part of the Mechanical Engineering Commons

Recommended Citation Kim, Namwon, "Multiphase flows in polymer microfluidic systems" (2009). LSU Doctoral Dissertations. 221. https://digitalcommons.lsu.edu/gradschool_dissertations/221

This Dissertation is brought to you for free and open access by the Graduate School at LSU Digital Commons. It has been accepted for inclusion in LSU Doctoral Dissertations by an authorized graduate school editor of LSU Digital Commons. For more information, please [email protected].

MULTIPHASE FLOWS IN POLYMER MICROFLUIDIC SYSTEMS

A Dissertation

Submitted to the Graduate Faculty of the Louisiana State University and Agricultural and Mechanical College in partial fulfillment of the requirements for the degree of Doctor of Philosophy

The Department of Mechanical Engineering

By Namwon Kim B.S., Kangwon National University, Korea, 1998. May 2009

To my parents,

Man Ki Kim and Keum Yeon Cho,

and lovely family

ii Acknowledgments

I would like to express my sincere gratitude to all who stood by me throughout this academic journey. The first two persons that come to my mind are my advisors, Dr. Michael C.

Murphy and Dr. Dimitris E. Nikitopoulos who advised me how to keep on the right track to this finish line, which also means another start line. Dr. Murphy is one of the most enthusiastic and patient people as a teacher and researcher I ever met. He always reminds me of a master who has a calm and gentle charisma. I have to say another sincere gratitude to Dr. Dimitris E.

Nikitopoulos, my co-advisor who let me through the maze of research with his tremendous academic knowledge, experiences, energies and Greek jokes. Additional gratitude is extended to

Dr. Steven A. Soper and Dr. Jin-Woo Choi for their guidance and discussion that made this research moving forward and Dr. Dandina N Rao who is a Dean’s representative.

There were so many people who inspired me hanging around me, not only by their academic knowledge, but also their attitudes about life. I gratefully acknowledge my research colleagues, Dr. Daniel S. Park and Jason Guy at the Center for Bio-Modular Multi-Scale

Systems (CBM2), Dr. Sunggook Park at ME, Dr. Yohannes Desta and Dr. Proyag Datta at Center for Advanced Microstructures and Devices (CAMD) for their invaluable academic advice and technical support in the fabrication of polymer microfluidic chips. I also thank all lab mates who got through weekly meeting together, Dr. Byoung Hee You, Dr. Pin-Chuan Chen and his family,

Taehyun Park, Tae Yoon Lee, Chetan Ramesh and Adam Cygan in Micro-Systems Engineering

Team (µSET). I also thank Dr. Sudheer Rani, Estelle Evans and Eamonn Walker in the

Microfluidics lab, Wonbae Lee and Dr. Subramanian Balamurugan in Chemisty, Jeong Tae Ok and Junseo Choi in Prof. Park’s lab, and Junpyo Hong, Dr. Won Kyo Jung, Jihwan Park, Sejong

Kim and Taewoo Park on the LSU tennis court.

iii Most of all, everything to my lovely family, I cannot express how I am grateful to their

love and support, my parents Man Ki Kim and Keum Yeon Cho, sister Hyekyung Kim, brother

Namhun Kim, parents-in-law Sung Ki Lee and Yeon Sook Choi, brother-in-law Jongmin Lee,

my companion for life, Eun Ju who took me over from my mom, my most precious son,

Kyungrae who has complete control of me, and may be another one.

iv Table of Contents

Acknowledgments ...... iii

List of Tables ...... vii

List of Figures ...... viii

List of Symbols ...... xv

Abstract ...... xvii

Chapter 1 Introduction ...... 1 1.1 Microfluidics and Bio-MEMS ...... 1 1.2 High Throughput Screening (HTS) with Microfluidics ...... 2 1.3 Objectives ...... 3 1.3.1 Fabrication of Microfluidic Polymer Chips ...... 3 1.3.2 Multi-Phase Flows in Microfluidic Devices ...... 4 1.4 Outline of Dissertation ...... 5

Chapter 2 Background ...... 7 2.1 Fundamental Equations in Microfluidics ...... 7 2.2 Dimensionless Numbers in Microfluidics ...... 8 2.3 Dispersion and Mixing in Multiphase Flow ...... 10 2.4 Thin Film in Multiphase Flow ...... 12 2.5 Literature Review ...... 15 2.5.1 Gas-Liquid Two-Phase Flow ...... 15 2.5.2 Liquid-Liquid Segmented Flow ...... 19

Chapter 3 Polymer Microfluidic Test Chips and Experimental Apparatus ...... 26 3.1 Introduction ...... 26 3.2 Preparation of Polymer Microfluidic Test Chips ...... 27 3.2.1 Direct Micromachining of Polymers ...... 27 3.2.2 Micromachining of the Brass Mold Insert ...... 29 3.2.3 X-Ray LIGA ...... 31 3.2.4 Hot Embossing of Thermoplastics ...... 33 3.2.5 Thermal Fusion Bonding and Interconnection ...... 35 3.3 Surface Roughness of Microchannels ...... 38 3.4 Experimental Apparatus ...... 42 3.4.1 Gas-Liquid Experiment Setup...... 42 3.4.2 Liquid-Liquid Experiment Setup ...... 44

Chapter 4 Gas-Liquid Two-Phase Flows in Microchannels ...... 47 4.1 Introduction ...... 47 4.1.1 Configurations of Microfluidic Test Chips ...... 47 4.2 Gas-Liquid Two-Phase Flow Regimes ...... 50 4.2.1 Capillary Bubble Flow ...... 50 4.2.2 Segmented Flow...... 51

v 4.2.3 Segmented-Annular Flow ...... 52 4.2.4 Annular Flow ...... 53 4.2.5 Dry Flow ...... 54 4.3 Gas-Liquid Two-Phase Flow Regime Maps ...... 55 4.4 Details of the Segmented Flow Regimes ...... 58 4.4.1 Image Processing ...... 60 4.4.2 Gas Bubble and Liquid Plug Lengths ...... 63 4.4.3 Regularity of Segmented Flow ...... 65 4.5 Pressure Drop in Microchannels ...... 72 4.5.1 Single Phase Frictional Pressure Drop ...... 72 4.5.2 Homogeneous Flow Model for the Two-Phase Flow Pressure Drop ...... 74 4.5.3 Separated Flow Model for the Two-Phase Flow Pressure Drop ...... 76 4.5.4 Measurements of Gas-Liquid Two-Phase Flow Pressure Drop ...... 78 4.6 Conclusions ...... 85

Chapter 5 Liquid-Liquid Segmented Flows in Microchannels ...... 86 5.1 Introduction ...... 86 5.1.1 Configurations of Microfluidic Test Chips ...... 87 5.2 Properties of Test Fluids ...... 89 5.2.1 Wettability and Surfactant ...... 90 5.2.2 Measurement of Viscosity ...... 93 5.2.3 Measurement of the Surface Tension ...... 94 5.3 Liquid-Liquid Segmented Flow Regimes ...... 96 5.3.1 Type I Chip with an Expansion Ratio of 16 ...... 98 5.3.2 Type II Chip with an Expansion Ratio of 4 ...... 101 5.3.3 Type III Chip with an Expansion Ratio of 2 ...... 101 5.4 Flow Maps ...... 103 5.5 Wetting of Dispersed Fluid ...... 106 5.6 Flow Velocity Measurement ...... 108 5.7 Liquid-Liquid Segmented Flow Pressure Drop ...... 110 5.8 Conclusions ...... 114

Chapter 6 Conclusions and Future Work ...... 117 6.1 Conclusions ...... 117 6.1.1 Microfabrication of Polymer Chips ...... 117 6.1.2 Experimental Study of Gas-Liquid Two-Phase Flow in Microchannels ...... 117 6.1.3 Experimental Study of Liquid-Liquid Segmented Flow in Microchannels ...... 120 6.2 Future Work ...... 121 6.2.1 High Throughput Bioassay Using Droplets and FCCS ...... 121 6.2.2 Measurement of Liquid Thin Film ...... 125

References ...... 127

Appendix A X-Ray LIGA Process ...... 138

Appendix B OPTIMASTM Macro ...... 155

Vita ...... 159

vi List of Tables

Table 2.1 Gas-liquid two-phase flow test parameters in literatures ...... 18

Table 2.2 Test parameters and applications of liquid-liquid segmented flows...... 25

Table 3.1 Properties and hot embossing parameters for PMMA and PC [96]...... 35

Table 3.2 Thermal fusion bonding conditions for PMMA and PC...... 38

Table 3.3 Surface roughness values of polymer chips...... 41

Table 3.4 Description of objectives performance ...... 43

Table 3.5 Specifications of fluorescence mirror units ...... 46

Table 4.1 Dimensions of the injection and test channels and the surface roughness of the side and bottom walls of the microchannels in the micro-milled and hot embossed chips...... 49

Table 4.2 Resolution of images acquired using objectives for different magnification and binning...... 61

Table 4.3 Parameter C in Lockhart-Martinelli correlation (Chisholm, 1967)...... 76

Table 4.4 Variables for the abscissa and ordinate used in Lockhart-Martinelli correlation and this work...... 79

Table 5.1 Characteristic dimensions of the injection and test channels ...... 89

Table 5.2 Properties of the dispersed and carrier fluids. DI-water and perfluorocarbo (FC3283) with 10% (v/v) nonionic surfactant (PFO, Perfluorooctanol) were used as test fluids...... 97

vii List of Figures

Figure 2.1 (a) Parabolic velocity profile by a pressure-driven flow and dispersion of molecules by concentration gradient (Taylor dispersion) results in high dispersion of molecules in the single phase flow. (b) In gas-liquid two-phase flow, dispersion of molecules occurs through the thin film and corner between gas bubble and channel walls. Recirculation of streamline enhances the mixing in liquid plug. (c) In liquid-liquid two-phase flow, molecules are encapsulated within dispersed fluid droplet or plug without dispersion. Recirculation of streamline exists both in the dispersed and carrier fluids...... 11

Figure 2.2 Side view of streamline in liquid plug (a) Recirculation of streamline for Ca < 0.7. (b) Complete bypass flow for Ca > 0.7, (Taylor, 1961 [36])...... 12

Figure 2.3 Cross-sectional view of a bubble in a square microchannel (a) Non- axisymmetric bubble (b) Axisymmetric bubble (Kolb and Cerro, 1991 [41])...... 13

Figure 2.4 Scaled liquid film thickness between gas bubble and channel walls in (a) a capillary tube and (b) square channel in function of capillary number (Ca) (Kreutzer et al., 2005 [39])...... 14

Figure 2.5 Instantaneous velocity vector and streamline in liquid plugs were obtained from μPIV and concentration field were acquire using fluorescence microscopy in (a) straight and (b) meandering microchannels of 400 μm side and 280 μm deep. (Gunther et al., 2005 [22]) ...... 15

Figure 2.6 (a) Merging of two droplets in different sizes into a plug and (b) symmetric and asymmetric splitting of plugs at a T-shaped junction depending on pressure in two branched outlet channels, (Song et al., 2003 [51])...... 20

Figure 2.7 Continuous segmented flows in a tube as an alternative to micro titer plate (a) reagent and substrate solution are dispensed by robotics in each well of titer plate (b) preloaded reagent plugs are mixed with substrate solution spontaneously, (Chen and Ismagilov, 2006 [10])...... 21

Figure 2.8 Screens of crystallization conditions for membrane protein with different concentration of precipitant in each plug, which showed different crystal patterns, (Li et al., 2006 [52])...... 21

Figure 2.9 (a) Loading of a series of reagents in parallel tube cartridges by controlling array of valves connected to vacuum (Linder et al. 2005 [11]) and (b) loading of a series of reagents in single input array and splitting into several output arrays using the repeated T-junctions (Adamson et al. 2006 [12])...... 23

Figure 3.1 (a) Kern MMP Micro-milling and drilling machine in the Center for BioModular Multi-Scale Systems (CBM2). (b) Working stage equipped with spindle and CCD camera with microscope for observation of machining...... 27

viii Figure 3.2 (a) Micromachined polycarbonate chips. (b), (c) Scanning electron microscope (SEM) images of Burrs along the 200um channels generated from machining on the polycarbonate...... 28

Figure 3.3 Micromachined channels on PMMA after ultrasonic agitation. (a) 50 µm depth × 50 µm width cross junction channel. (b) concave curvature caused by the tool. (c) 200 µm depth × 200 µm width serpentine channels...... 29

Figure 3.4 (a) Photograph of micromachined brass mold insert. SEM pictures of brass mold insert. (b) 50 µm width × 50 µm height cross shaped brass mold structure. (c) 200 µm width × 200 µm height curved brass mold structure showing the striation on the surface. (d) hot-embossed PMMA corresponding to (b). (e) hot- embossed PMMA corresponding (c)...... 30

Figure 3.5 Fabrication of polymer microchip using LIGA: (a) X-ray exposure of CQ grade PMMA photoresist bonded on a stainless steel substrate with X-ray mask; (b) Developing of PMMA using GG developing solution; (c) Over-electroplating of nickel in the developed PMMA master; (d) Laser welding between machined nickel mold and grooved stainless steel substrate; (e) Hot embossing of thermoplastic with fabricated mold insert; and (f) Microchip enclosed by the same polymer film by thermal bonding...... 32

Figure 3.6 Hot embossing machine at CAMD. (a) HEX 02 (JENOPTIK Mikrotechnik, Jena, Germany). (b) Upper plate for mold insert. (c) Bottom plate for polymer to be embossed...... 34

Figure 3.7 Thermal bonding jig...... 36

Figure 3.8 (a) NanoportTM glued on polymer chips used in gas-liquid two-phase flow experiments. (b) Intersertion of PEEK tubes into polymer chips used in the liquid-liquid segmented flow experiments...... 38

Figure 3.9 SEM images of the sidewalls of the microchannels in (a) a direct micromachined PMMA chip, (b) a chip hot embossed PMMA with a micromachined brass mold insert; and (c) a chip hot embossed with a LIGA nickel mold insert. (d), (e) and (f) are optical profiler images corresponding to (a), (b) and (c)...... 40

Figure 3.10 SEM images of the bottom of a chamber in (a) a direct micromachined PMMA chip, (b) a chip hot embossed in PMMA with a micromachined brass mold insert; and (c) a chip hot embossed in PMMA with a LIGA nickel mold insert. (d), (e) and (f) are optical profiler images corresponding to (a), (b) and (c)...... 40

Figure 3.11 Surface roughness profiles of a polymer microfluidic chip. (a) Surface roughness profiles from the bottom of the chamber. (b) Surface roughness profiles from the sidewall of the microchannel...... 41

Figure 3.12 Schematic of the experimental apparatus for gas-liquid two-phase flow ...... 42

ix Figure 3.13 Schematic of the experimental apparatus for liquid-liquid segmented flow...... 45

Figure 4.1 (a) Schematic of the microchannels for the PMMA chip hot embossed from a micro-milled brass mold insert. The chip has filleted corners due to the finite drill bit radius (r = 100 µm). (b) Photograph of the fabricated PMMA chip...... 48

Figure 4.2 Capillary bubbly flow (CB), [Lb/wt < 1]: gas superficial velocity (JG) ≈ 0.014 m/s, liquid superficial velocity (JL) ≈ 0.093 m/s and the liquid volumetric flow ratio (βL) ≈ 0.87...... 50

Figure 4.3 Segmented flow (S): (a) Segmented-1 (S1), [(Lb-wt)/(Lp+wt) < 1, Lb/wt < 5], JG ≈ 0.069 m/s, JL ≈ 0.064 m/s and βL ≈ 0.43, (b) Segmented-2 (S2), [(Lb- wt)/(Lp+wt) > 1, Lb/wt < 5], JG ≈ 0.417 m/s, JL ≈ 0.0742 m/s and βL ≈ 0.15, (c) Segmented-3 (S3), [(Lb-wt)/(Lp+wt) > 1, Lb/w > 5], JG ≈ 0.638 m/s, JL ≈ 0.058 m/s and βL ≈ 0.068...... 51

Figure 4.4 Segmented-Annular flow (SA): gas superficial velocity (JG) ≈ 1.72 m/s, liquid superficial velocity (JL) ≈ 0.084 m/s and the liquid volumetric flow ratio (βL) ≈ 0.047...... 52

Figure 4.5 Annular flow (A): gas superficial velocity (JG) ≈ 4.19 m/s, liquid superficial velocity (JL) ≈ 0.057 m/s and the liquid volumetric flow ratio (βL) ≈ 0.014...... 53

Figure 4.6 Dry flow (D): gas superficial velocity (JG) ≈ 12.13 m/s, liquid superficial velocity (jL) ≈ 0.002 m/s and the liquid volumetric flow ratio (βL) ≈ 0.0016...... 54

Figure 4.7 Gas-liquid two-phase flow map with regime separation lines for the flow in the serpentine test microchannels of the Type I AR=1 and Type II AR=2 chips...... 55

Figure 4.8 Gas (JG) and liquid (JL) superficial velocities in term of liquid volumetric flow ratio (βL). Liquid superficial velocity (JL) was affected by gas superficial velocity (JG) due to the change of hydraulic resistance...... 56

Figure 4.9 Two-phase flow map for flow in serpentine test microchannels with different aspect ratios (AR) of mico-milled chips, (Dh=200 μm, +/-3%). Regime separation lines from previous works of Cubaud and Ho [19], Günther and Jensen [102], and Triplett et al. [104] (B-Bubbly, W-Wedge, S-Slug, A- Annular, WA-Wavy/Annular, AD-Annular/Dry, and D-Dry)...... 57

Figure 4.10 Image processing of 8-bit grayscale digital images to get length of gas bubble and liquid plug. (a) one of raw images from Segmented flow regime; (b) Clear field image to remove channel edge shown in raw image; (c) inverted image dividing the raw image by clear field image and inverting pixel value; (d) detecting gas bubbles by filling the confined gas bubble area and acquiring geometry including the length of the major axis, centroid, and number of bubbles; (e) mask image to get the length of liquid plug by adding the image to image (d); (f) detecting the liquid plug and extracting the length of the liquid plug...... 62

x Figure 4.11 Scaled gas bubble and liquid plug lengths with respect to (a) the liquid volumetric flow ratio and (b) Capillary number for the Capillary bubbly and all Segmented flows in the microchannel of the hot embossed chips. Distribution of gas bubble length corresponding a’ to e’ are shown in Figure 4.12 (g)...... 64

Figure 4.12 Representative distributions of the gas bubble length and images of air-water two-phase flows in the PMMA serpentine microchannels of hot embossed chip with Dh=200μm (nominal). Data points shown in (g), from (a’) to (e’), are corresponding to those in Figure 4.12 (a). (a) Capillary bubbly, βL=0.804, number of sample (bubble): 6,001, Range: 19.9µm, Mean: 152.61µm, Coefficient of variation (CV): 1.39%; (b) Segmented-1, βL=0.433, number of sample: 8,948, Range: 39µm, Mean: 235.24µm, CV: 2.76%; (c) Segmented-2, βL=0.26, number of sample: 6,897, Range: 64µm, Mean: 326.76µm, CV: 3.91%; (d) Segmented-2, βL=0.12, number of sample: 2,833, Range: 880.8µm, Mean: 816.88µm, CV: 15.5%; (e) Segmented-3, βL=0.068, number of sample: 1,246, Range: 2,838µm, Mean: 1,291.66µm, CV: 30.28%; (f) Observation region...... 65

Figure 4.13 Images of regular segmented air-water two-phase flow at comparable bulk flow conditions in PMMA serpentine microchannel (micro-milled chip with 3 aspect ratio test channels) with Dh=200 µm (nominal) for (a) AR=1, Segmented-2, βL=0.25, JL=46 mm/s; (b) AR=2, Segmented-2, βL=0.266, JL=43.4 mm/s; (c) AR=3, Segmented-3, βL=0.292, JL=46.2 mm/s...... 66

Figure 4.14 (a) Gas bubble and (b) liquid plug length distributions and their dependence on channel aspect ratio for low liquid superficial velocities. (AR=1: βL=0.304, JL=47.4 mm/s; AR=2: βL=0.313, JL=44 mm/s; AR=3: βL=0.339, JL=46 mm/s). .... 67

Figure 4.15 Illustration of two mechanisms responsible for irregularity in Segmented flow regime: Instability of the segmented flow resulting in coalescence (a) βL =0.671, JL =54.7 mm/s, (b) associated bubble length distribution, and Irregular injection at the exit from the mixing section (c) βL=0.18, JL=58.8 mm/s (d) associated bubble length distribution...... 68

Figure 4.16 (a) Regular and (b) irregular segmented flow through coalescence at the same bulk flow conditions AR=1, βL=0.415, JL=57.7 mm/s...... 70

Figure 4.17 Two-phase frictional multiplier 2 in terms of Lockhart-Martinelli parameter (X)...... 77

Figure 4.18 Two-phase frictional multiplier ( 2 ) in terms of Lockhart-Martinelli parameter () with constant C=1.39 for AR=1 test channels. CB: Capillary Bubbly, S1: Segmented-1, S2: Segmented-2, S3: Segmented-3, SA: Segmented-Annular, A: Annular and D: Dry flows...... 78

Figure 4.19 Scaled two-phase pressure drop as a function of liquid volumetric flow ratio (βL) for (a) all flow regimes and (b) details of Segmented flow regime...... 81

xi Figure 4.20 Scaled two-phase pressure drop as a function of Capillary number (Ca) for (a) all flow regimes and (b) details of Segmented flow regime...... 82

Figure 4.21 The number of gas bubbles present in the channel between the two pressure ports with respect to the liquid volumetric flow ratio for the Capillary bubbly and all Segmented flows in the microchannel of hot embossed chip...... 83

Figure 5.1 Schematics of the hot embossed polycarbonate test chips. (a) Type I with an expansion ratio from the injection to the test channel of 1:16 (b) Type II with an expansion ratio of 1:4 and (c) Type III with an expansion ratio of 1:2...... 88

Figure 5.2 Wettability of (a) FC 3283 + PFO (10% v/v) solution on PMMA (Complete wetting) (b) FC 3283 + PFO (10% v/v) solution on PC (complete wetting) (c) the deionized water on PMMA (Contact angle ≈ 69°) (d) deionized water on PC (Contact angle ≈ 85°)...... 91

Figure 5.3 (a) Young’s relation of sessile drop under the static condition, (b) nonionic fluoro-soluble surfactant (1H, 1H, 2H, 2H- perfluorooctanol, CF3(CF2)5(CH2)2OH) (c) the present of surfactant decrease γSO and γWO (d) surface treatment increase γSW and decreased γSO...... 92

Figure 5.4 Distinct dynamic contact angles between the carrier fluid and channel walls in liquid-liquid segmented flow (a) with and (b) without surfactant in carrier fluid at the 20 µl/min carrier fluid volumetric flow rate (QC) 3 µl/min and dispersed fluid volumetric flow rate (QD)...... 93

Figure 5.5 Callibrated Cannon-Fenske Routine Viscometers according to ASTM D 445 and ISO 3104...... 94

Figure 5.6 (a) Surface tension and interfacial measurement system (FTA125) (b) Water pendant drop suspended in air with variables representing drop shape...... 95

Figure 5.7 FC 3283/deionized water, γFW = 54.15 ± 0.13 mN/m, (b) FC 3283 + PFO (5% v/v) solution/deionized water, γFW = 14.79 ± 0.03 mN/m, (c) FC 3283 + PFO (10% v/v) solution/deionized water, γFW = 13.49 ± 0.33 mN/m, and (d) FC 3283 + PFO (20% v/v) solution/deionized water, γFW = 12.25 ± 0.42 mN/m...... 96

Figure 5.8 Density of the carrier fluid, the interfacial force between the carrier and dispersed fluids, and the dynamic viscosity of the carrier fluid as a function of surfactant (PFO, Perfluorooctanol) volumetric concentration (% v/v)...... 97

Figure 5.9 Liquid-liquid segmented flow regimes of the 50 µm width × 50 µm depth injection channel chip (Type I chip with an expansion ratio of 16) under white field illumination (a) Droplet flow in the injection channel (4X objective) with homogeneous carrier fluid volumetric flow ratio, βC ≈ 0.93 under white field illumination (b) Droplet flow in the test channel (10X objective) with βC ≈ 0.95 under laser illumination (c) Droplet flow, βC ≈ 0.74 (d) Irregular Segmented flow, βC ≈ 0.75 (e) scattered Droplet flow, βC ≈ 0.5 (f) Irregular Segmented flow, βC ≈ 0.5 (g) Plug flow, βC ≈ 0.37 (h) Plug flow, βC ≈ 0.37...... 99

xii Figure 5.10 Liquid-liquid segmented flow regimes of the 50 µm width × 200 µm depth injection channel chip (Type II chip with an expansion ratio of 4) under white field illumination (a) Droplet flow at the injection channel (4X objective) with homogeneous carrier fluid volumetric flow ratio, βC ≈ 0.87 (b) Droplet flow in the test channel (2X objective) with βC ≈ 0.87 (c) Irregular Segmented flow, βC ≈ 0.69 (d) Irregular Segmented flow, βC ≈ 0.67 (e) Plug flow, βC ≈ 0.4 (f) Plug flow, βC ≈ 0.4 (g) Plug flow, βC ≈ 0.11 (h) Plug flow, βC ≈ 0.11...... 100

Figure 5.11 Liquid-liquid segmented flow regimes of the 100 µm width × 200 µm depth injection channel chip (Type II chip with an expansion ratio of 2) under white field illumination (a) Plug flow at the cross junction area (4X objective) at carrier fluid volumetric flow ratio, βC ≈ 0.83, (b) Plug flow at the test channel, βC ≈ 0.83, (c) Plug flow, βC ≈ 0.5, and (d) Plug flow, βC ≈ 0.2...... 102

Figure 5.12 Liquid-liquid segmented flow regime map and transition lines between regimes observed from the test channel of Type I chip with expansion ratio (ER) 16. (●: Droplet flow, ▼: Irregular segmented flow, and ■: Plug flow) ...... 104

Figure 5.13 Liquid-liquid segmented flow regime map and transition lines between regimes observed from the test channel of Type II chip with expansion ratio (ER) 4. (●: Droplet flow, ▼: Irregular segmented flow, and ■: Plug flow) ...... 104

Figure 5.14 Measurement of the dispersed and carrier fluid lengths in terms of carrier fluid volumetric flow ratio from Type II chip...... 105

Figure 5.15 Wetting of dispersed fluid on the channel surface (a) in Droplet flow, and (b) Plug flow. (c) and (d) different segmented flow regimes observed under the same flow conditions in the same test chip due to wetted patches in the injection channel, QC = 20 µm/min, QD = 3 µm/min and βC ≈ 0.87...... 107

Figure 5.16 Overlap of the consecutive images taken with the double pulsed laser with a 2.5 msec pulse separation for the (a) Plug flow and (b) Droplet flow regimes (c) measured velocity (VD) on the Type I chip was scaled by sum of superficial velocities of the disperses and carrier fluids (J = JC + JD)...... 109

Figure 5.17 (a) Liquid-liquid segmented flow pressure drops and (b) scaled segmented flow pressure drops by single liquid flow pressure drops as a function of the carrier fluid volumetric flow ratio...... 111

Figure 5.18 Experimental measurement of friction factor, f, in terms of Reynolds number for the carrier fluid single phase flow pressure drop...... 113

Figure 5.19 Measured two-phase friction multiplier data in terms of Lockhart-Martinelli parameter (b) comparison of measured and predicted liquid-liquid segmented flow pressure drop with C=4.63...... 115

Figure 6.1 Monitoring of enzyme activity in droplet for high throughput bioassay using fluorescence cross-correlation spectroscopy (FCCS)...... 122

xiii Figure 6.2 Schematic illustration of the fluorescence cross-correlation spectroscopy (FCCS) setup using dual laser sources, 532nm and 780 nm diode lasers, (Image of courtesy of Wonbae Lee – Department of Chemistry, LSU)...... 123

Figure 6.3 (a) Hot embossed polycarbonate chip with 50 µm × 50 µm injection channel and 200 µm × 200 µm test channel. Fluorescent intensity was detected from the indicated point in the test channel (b) microscopic image and (c) fluorescent signal of droplet flow in test channel with dispersed fluid flow rate, QD = 1 µl/min and carrier fluid flow rate, QC = 20 µl/min (d) microscopic image of droplet flow in test channel with dispersed fluid flow rate, QD = 1.8 µl/min and carrier fluid flow rate, QC = 20 µl/min and (e) fluorescent signal of droplet flow in test channel with dispersed fluid flow rate, QD = 2 µl/min and carrier fluid flow rate, QC = 20 µl/min ...... 124

Figure 6.4 (a) Experimental scheme for the detection and measurement of liquid thin film between the dispersed liquid plug and channel wall using fluorescence cross- correlation spectroscopy (b) Fluorescence excitation and emission spectra of fluorescent microsphere and quantum dot (Image courtesy of Invitrogen)...... 126

xiv List of Symbols

A Cross-sectional area, m2

A Acceleration vector, m/s2

AR Aspect ratio, d/w

Ca Capillary number, µV/σ

D Mass diffusion coefficient, m2/s

d Channel depth, m

Dh Hydraulic diameter, m

F Force vector, N

f Friction factor

f Force density vector

J Superficial velocity, m/s

L Characteristic length, m

m mass, kg

P Pressure, Pa

Pe Péclet number, LV/D

Q Volumetric flow rate, ml/min

R Surface roughness, m

r Radius, m

Re Reynolds number, ρVL/µ

RMS Root mean square

Tg Glass transition temperature, °C

t Time, sec

u Velocity vector

xv V Velocity, m/s

V Volume, m3

w Channel width, m

We Weber number, ρV2L/γ

X Lockhart-Martinelli parameter z mixing length, m

Z Unit channel length, m

Greek symbols

β Volumetric flow ratio

γ Interfacial force, N/m

δ Liquid film thickness, m

μ Dynamic viscosity, Pa·s

ρ Density, kg/m3

Lockhart-Martinelli multiplier

Subscripts

b Gas bubble

G Gas phase

L Liquid phase

l Liquid plug

P Pulse separation

S Single phase flow

TP Two-phase flow

xvi Abstract

Continuous delivery of segmented reagents using pressure-driven multiphase flow in

microchannels is a promising technology for high throughput microfluidic bioassays. Separation and encapsulation of the target reagents with another inert fluid provide many advantages over single phase flow in microfluidic applications of biotechnology. In order to achieve these advantages and control these multiphase flows, it is necessary to understand their generation and transport characteristics as influenced by geometrical miniaturization, channel wall properties,

the effects of surfactants and operating conditions.

For gas-liquid two-phase flow, dry air and deionized water were driven into hot

embossed PMMA microchannels with 200 μm square test microchannels. Flow regimes, flow

maps and the lengths of the gas bubbles and liquid plugs in terms of the liquid volumetric flow

ratio (βL) were determined. Continuous generation of regular segmented flow was also discussed.

Three sub-regimes of the Segmented flow were identified based on the statistical phase length

scales observed over a substantial test channel length.

For the liquid-liquid segmented flow, deionized water and perfluorocarbon with a

surfactant were used as test fluids in the hot embossed polycarbonate microchannels. The effects

of three expansion ratios from the injection to the test channels of 2, 4, and 16 were investigated

comparing the flow regimes, transitions and maps in terms of a fixed carrier fluid volumetric

flow ratio. The length of the dispersed fluids and the distance between consecutive droplets or

plugs in terms of the carrier fluid volumetric flow ratio (βC) were determined. Velocities of the

dispersed droplets and plugs were measured using double-pulsed laser illumination and were found to be 1.46 ± 0.08 and 1.25 ± 0.05 times faster than the superficial velocity of the

segmented flow, respectively.

xvii The multiphase flow pressure drops were measured for all of the flow regimes in gas- liquid two-phase and liquid-liquid segmented flows. Each flow regime identified on the basis of topological observations, including the length scale of each fluid phase and the number of the gas bubbles or dispersed droplets in unit length with respect to the volumetric flow ratio, was associated with different trends in the pressure drop variation.

xviii Chapter 1 Introduction

1.1 Microfluidics and Bio-MEMS

There are increasing demands for highly developed analytical systems that allow

biologists and chemists to save resources and time while retaining reliable outputs. Miniaturized analytical systems called Lab-on-a-Chip (LOC) or Micro Total Analysis Systems (µTAS) have emerged as the leading contender for meeting these needs. In addition to the reduced reagent volume and process time consumed in the analysis, miniaturization of reaction systems improves the molecular diffusion and heat transport without changing the nature of the molecular reactions due to the increased surface to volume ratio [1]. As a result of the development of these miniaturized systems, portable and handheld bio-analytical devices shrunken from a bench-top size device are expected in the near future. The miniaturized devices could be deployed in the field wherever instantaneous and fast responses are required. This is a similar path of evolution as for handheld computers which originated from the room-sized vacuum-tube computer. As the integrated circuits are the main components of computers, fabricated functional microchannels are the key parts of the Lab on a Chip. Accumulated data and information on fabrication methods of integrated circuits in nano-scale and digital signal processing made the splendid evolution of the computer feasible. Even though the development of LOC is at its beginning, there are numerous researches on the fabrication of the integrated chip and independent functions for microfluidic devices such as pumps, valves, mixers, sensors and controllers. There are unique physical phenomena induced by the geometrical miniaturization of bio-analytical systems.

Microfluidics is the subject dealing with all the physical phenomena induced by the geometrical miniaturization of fluidic systems. There are increasing numbers of fundamental studies on the physics arising from working fluids in microchannels. All of these efforts are leading to the continuous evolution of the Lab-on-a-Chip.

1 1.2 High Throughput Screening (HTS) with Microfluidics

Conventionally, biologists and chemists perform their experiments attempting to find a designated result by conducting systematic experiments with biochemical reagents on the order of liters or milliliters through sample preparation, delivery of the chemicals to the glassware, labeling, mixing, and recording results and repeating the processes for screening. While the purpose of the screening process has not been changed, technologies are continuously evolving to achieve higher throughput screening of tens or hundreds of thousands of bioassays at the molecular level on an efficient time scale. Traditional glassware with hand-written labels was replaced by 96-well micro titer plates and computer tracking of identities in the separate wells, which are essential parts of the High Throughput Screening (HTS). Micro titer plates are polymer plates with wells commonly used in analytical experiments, like biochemical screening processes. The standard defining micro plates was initially published in 1995 and updated in

2004 by the American National Standards Institute (ANSI) on behalf of The Society for

Biomolecular Sciences (SBS) [2]. At the initial stage on usage of the titer plate, reagents and substrates on the order of milliliters were dispensed to each well using micro-pipettes by operators. However, with the development of automation and computer technology, thousands of biochemical substances are assayed in a rapid process within a limited time. These automation systems are equipped with robot arms that dispense precise amounts of aqueous solution into each well and convey the plate among process equipment [3, 4]. However, there are still highly increasing demands for ultra high throughput screening methods based on microfluidics, especially in genetic, proteomic, pharmacological, and forensic analysis. Microfabrication processes originating in the semiconductor industry make it possible to integrate more than 9,600 wells in a single plate [5] and the high density well format improves high throughput screening process [6]. Microdevices for the polymerase chain reaction (PCR), ligase detection reaction

2 (LDR), and capillary electrophoresis (CE) are also integrated in a single plate for the high throughput genetic analysis. In spite of these advances in automation systems and fabrication techniques, precise delivery and manipulation of minimized volumes of reagents in microfluidic networks raises another research issue. Especially, in the case of repeating screening processes with high cost and limited quantities of reagent, it is critical to minimize the amount of reagent consumed in each screening step. However, current robotic dispensing systems have some limitations in handling sub-microliter liquid volumes which tend to evaporate easily from the open wells [6, 7]. Microfluidic polymer chips provide sealed channels that prevent evaporation of reagents during the reaction process. Once the liquid sample is introduced into the channel, manipulation of liquid is not affected by the external dispensing system [1].

The simplest way to carry out analysis and synthesis reactions in microchannels is continuous flow, in which a single phase aqueous reagent flows in a long microchannel for a specific reaction [8]. In order to increase high throughput capacity, single phase reagents can be separated by another fluid into a series of plugs with distinct chemical entities [9, 10].

Furthermore, parallel channels with consecutive arrays of plugs can be integrated in microfluidic chips for massive parallelization [11, 12]. Analytical screening and synthesis can be also performed within high density separate wells or devices integrated on the chips in a format similar to a micro titer plate [13-15].

1.3 Objectives

1.3.1 Fabrication of Microfluidic Polymer Chips

Microfluidics came to play a major role in the field of biomedical applications usually requiring a number of repeated tests for reliable results of biochemical reactions. While the application began with platform based on silicon with the help of integrated circuit fabrication technology, thermoforming of polymers allows cost-effective mass production of the 3 microfluidic platforms designed for the biomedical applications. Additionally, each polymer chip integrated with a function can be modulated in serial or parallel to function as a system.

Ultimately, this modulation of each polymer chip enables integration of all of the functions of current analytical devices into a single microfluidic chip.

Reliable production of low cost disposable polymer chips is the one of the objectives of this work. Using fabrication facilities available at the Center for Bio-Modular Multi-Scale

Systems (CBM2) and Center for Advanced Microstructures and Devices (CAMD) in LSU, three

different types of polymer microfluidic chips were prepared. Direct milling-machining of

microchannels in the polymer provide rapid prototyped platforms for functional device

evaluation. Mold inserts fabricated by micro-milling and the UV- or X-ray LIGA processes

enabled polymer chips to be replicated in large quantities through thermoforming. Process

parameters and characteristic investigations of chips from each fabrication method were

necessary to adapt proper polymer chips for specific applications.

1.3.2 Multi-Phase Flows in Microfluidic Devices

Multiphase flow in microfluidic systems is realizing its potential for enhancing the

performance of miniaturized biochemical reaction systems as a promising reagent delivery method [16-18]. The common approach for delivering a liquid reagent for the performance of analytical and synthetic reactions in microchannels is a continuous, single phase liquid flow. In this case, a part of the liquid reagent is used to drive the medium, not strictly for the reactions, since the whole channel must be filled. Segmenting the liquid reagent into packets with an inert second fluid reduces the volume of liquid reagent of interest and increases the surface to volume ratio locally, so that biochemical reactions in the segmented liquid reagents are accelerated more within a limited microchannel volume. Segments of liquid flow with another fluid also induce recirculation of molecules inside the liquid plugs, which reduces the mixing times that are

4 otherwise restricted due to diffusion by the low Reynolds number laminar flow characteristics of

the microchannels.

Studies have examined gas-liquid and liquid-liquid two-phase flows in microchannels

etched in silicon with glass cover plates [17, 19, 20], cast in PDMS using soft lithography [21-

23], or conventional capillary tubes [12, 24], which were either partially or entirely molecularly

smooth with partially wetting walls. In order to design polymer microfluidic bio-analytical systems effectively and predictably using multiphase flows, including gas-liquid two-phase and liquid-liquid segmented flows, it is necessary to understand fundamental aspects of their behavior. An experimental investigation of multiphase flows in polymer microfluidic channels was carried out. Details of multiphase flow regimes and maps, length measures of the dispersed

and carrier fluid segments, and velocities measurements of dispersed fluid parcels were performed with particular interest in segmented flows and their structure and regularity.

Correlations of the multiphase flow pressure drops with the flow topology were also made.

While the microchannels used in the miniaturized bio-analytical devices improve heat transfer and mixing efficiency due to the higher surface to volume ratios, the pressure drop increases due to the micro-scale cross-sectional area of the microchannel. By taking advantage of the improved physical characteristics of the microchannels, understanding and prediction of pressure drop are also required prior to employing the segmented flow for the practical application in bio- analytical system.

1.4 Outline of Dissertation

This dissertation contains six chapters and appendices. The followed is a brief overview of the contents of each chapter.

Chapter 1 introduces the microfluidics and multiphase flow in polymer microchannels as a main theme and also provides the motivation and objectives of this research. Application of

5 multiphase flows in microfluidic polymer chip for reagent delivery in high throughput analysis in biotechnology is proposed. Chapter 2 provides fundamentals of microfluidics and multiphase flow in microchannels along with review of previous and current research on gas-liquid two- phase and liquid-liquid segmented flows to help a clear understanding of the chapters to come.

Chapter 3 treats fabrication steps for the preparation of microfluidic polymer test chips including three different methods such as direct milling-machining of channels on the polymer chips, machining of brass mold inserts for thermoforming, and X-ray LIGA mold inserts for thermoforming. The experimental apparatus and data acquisition methods are also described. In

Chapter 4, experimental studies of gas-liquid two-phase flow in microchannels are demonstrated.

Gas-liquid two-phase flow regimes, maps, irregularity, specifics of segmented flow and pressure drops in microchannel are discussed. Experimental study of liquid-liquid segmented flow in microchannel is included in Chapter 5. Flow regimes and maps are determined dealing with effects of the injection to test channel expansion ratio on flow regimes. Velocity and pressure drop of segmented flow are investigated. Chapter 6 summarizes results and proposes future work as a potential contribution on this academic field. Appendix A includes details of X-ray LIGA fabrication process for mold insert. Additionally, the image processing code, OptimasTM macro, for the automation of image processing is shown in Appendix B.

6 Chapter 2 Background

Reviews of previous and current research related to gas-liquid two-phase and liquid-

liquid segmented flows and information on the physical fundamentals of multiphase flows in

microchannels are provided in this chapter. Various practical applications of multiphase flows in microfluidic systems are also reviewed.

2.1 Fundamental Equations in Microfluidics

The basic equations governing fluid motion in microchannels under the continuum assumption are laws for conservation mass (continuity) and momentum (Newton’s second law).

Over the characteristic length scale of roughly 1μm, the fluid can be expected to behave as a continuum [25].

In Lagrangian terms, the law of conservation of mass is

V (2.1) where m is the mass, ρ is the density and V is the volume. In Eulerian terms, the law is

· 0 (2.2)

Conservation of momentum, known as Newton’s second law, relates the applied forces

(F) to the resulting acceleration of a mass m as;

(2.3)

In fluidic systems, forces (F) applied per unit volume can be written in Eulerian terms,

(2.4) where f is the force density and divided into surface forces and body forces. The surface forces are applied on the sides of the element by external viscous stresses. By substituting viscous stress

7 relations into Newton’s law and assuming that the viscosity and density are constant (i.e.,

incompressible Newtonian fluid), the Navier-Stokes equation can be written in the general form;

· (2.5) where terms on the left are for the inertial acceleration and terms on the right for body and surface forces. In microfluidics, the inertial forces are negligible compared to the viscous forces so the second term in the left hand side of the Navier-Stoke equation can be eliminated leaving the Stokes equation;

(2.6)

2.2 Dimensionless Numbers in Microfluidics

Fluid behaviors are subject to the physical properties and interactions among the forces exerted on fluids such as inertial force (density), viscous force (viscosity), gravitational force

(gravity), interfacial force (surface tension), convection (fluid motion), and diffusion (molecular motion). Dimensionless numbers explain the competition between the applied forces and give a sense of the relative contributions of each force in determining the characteristics of fluid flow in the microchannel. The dimensionless number used most frequently in characterizing microfluidic is the Reynolds number (Re), which is the ratio of inertial to viscous forces;

(2.7) where ρ is the density, V is the average flow velocity, L is the characteristic length scale, and the

μ is the viscosity. With typical gases and liquids in microchannels with characteristic lengths

8 from 10 to 200 μm, Reynolds numbers (Re) range between 10-3 and 10. These low values of the

Reynolds number show that viscous forces dominate inertial forces and result in laminar flow in

microchannels. The Stokes equation (Equation 2.6) results from eliminating the nonlinear terms

in the Navier-Stokes equation (Equation 2.5).

The small Re allows simple calculations using the Stokes equation without nonlinear term.

However, as a cost for this convenience, long mixing time is required because the mixing is

achieved by diffusion alone resulting from the laminar flow in microchannels. While slower

mixing is preferred in some applications such as cell sorting [26, 27] and particle separations

[28-30], faster mixing is usually desirable in biochemical reactions to mix different solutions

[31-33]. For this purpose, the Péclet number (Pe) gives an estimate of the mixing length, which is defined as the length in the flow direction after which the fluid composition over all positions of a channel cross-section deviated by no more than 1% of the equilibrium composition [34]. The

Pe expresses the relative importance of convection to diffusion;

~ (2.8) where L is the characteristic length scale, V is the average flow velocity, D is the mass diffusion coefficient, and z is the mixing length. The full mixing of a fluid depends linearly on Pe.

In multiphase flow, the most meaningful dimensionless number is the Capillary number

(Ca) expressing interplay between viscous and surface forces and defined as;

(2.9) where μ is the viscosity , V is the average flow velocity, and γ is the interfacial tension between fluids. Viscous forces work tangentially at the interface between fluids to elongate and drag it and interfacial forces work normal to the interface of fluids to minimize the interfacial area. The

9 competition between two forces governs the characteristic behaviors of multiphase flows in

microchannels.

2.3 Dispersion and Mixing in Multiphase Flow

In biochemical microfluidic reactors, dissolved molecules or suspended particles of

interest act like the fluid and are governed by the fluid motion. The efficiency of microreactors using pressure-driven single phase flow is limited by slow diffusive mixing and the broad residence time distribution (RTD) [see Figure 2.1 (a)]. The laminar flow characteristics in microchannel limited mixing rate even with an aid of Taylor dispersion (i.e. convection- diffusion) enhancing the mixing compared to pure diffusive mixing. The parabolic velocity profile of the single phase flow driven by pressure gradients under no-slip condition (Poiseuille flow) produces broad residence time distribution (RTD). However, separating a flow with another immiscible fluid prevents molecules from dispersing along the channels and induces the recirculation of streamlines in the separated fluid plugs which enhances mixing of molecules. In

the case of gas-liquid two-phase flow [see Figure 2.1 (b)], target molecules distribute within the

liquid plug and mixing is intensified by the circulating streamline. However, dispersion across

neighboring liquid plugs, although less than in single phase flow, is still occurs through liquid

film and corner flows between the gas bubble and channel walls. Muradoglu et al. [35] studied the effect of Péclet number (Pe), Capillary number (Ca) and segment sizes on the axial dispersion through the film in gas-liquid two-phase segmented flow computationally. For liquid- liquid segmented flow [see Figure 2.1 (c)], target molecules are confined in the dispersed fluid plugs. There is also recirculation within the dispersed fluid plugs producing higher mixing efficiency. If there is no wetting of the channel walls by the dispersed fluids, the molecules encapsulated in the dispersed fluid plugs are preserved without adsorptions on the channel wall, which could cause cross-contamination between consecutive dispersed plugs. This is the one of

the reasons that liquid-liquid segmented flow is a potential candidate for use in high throughput

screening.

As a pioneer in the field of two-phase flows, Taylor (1961) [36] investigated elongated

gas bubbles separating a liquid flow in a small channel (i.e., Taylor flow). The separation of the liquid flow by gas bubble induces circulating streamline in the liquid plugs. The circulating streamlines are separated by a thin film attached to the channel wall as shown in Figure 2.2 and

11 two flow patterns, bypass flow and recirculation of the streamline, are functions of the Capillary

number (Ca). When the capillary number (Ca) is smaller than 0.7, there is recirculation of

streamlines in the liquid plugs. Otherwise, for Ca > 0.7, flow passes through the liquid film and

corner without recirculation. Günther et al. [21, 22] studied the recirculation in liquid plugs flowing along straight and meandering channels experimentally using micro particle image velocimetry (µPIV) and fluorescence microscopy. Waelchli and Rohr [37] also showed that the recirculation inside the liquid plug was affected by the interfacial force rather than the viscosity of liquid, by investigating the distribution of absolute and relative velocities inside several plugs of different liquids including water, ethanol and glycerol using µPIV.

2.4 Thin Film in Multiphase Flow

Another important characteristic of multiphase flow is the presence of the thin film separating the dispersed fluid from the channel wall. A renowned lubrication theory for two- phase flow was developed by Bretherton [38] to show that the film of thickness is a function of the dispersed fluid velocity. Lubrication theory related the Laplace pressure difference across the gas-liquid interface and the viscous force in the liquid film using the full Navier-Stokes

(a) (b)

Figure 2.2 Side view of streamline in liquid plug (a) Recirculation of streamline for Ca < 0.7. (b) Complete bypass flow for Ca > 0.7, (Taylor, 1961 [36]).

12 equations and resulted in a scaling rule, Equation 2.10, in terms of the Capillary number (Ca)

[39].

0.66⁄ (2.10)

The correlation developed by Aussilous and Quére [see Equation 2.11] taking account of

inertial effects showed thicker liquid films than the prediction of Bretherton’s lubrication theory

[40].

0.66⁄ (2.11) 1 3.33⁄

Kreutzer et al. [39] derived a correlation (Equation 2.12) between the bubble diameter,

scaled by the channel diameter in the diagonal direction ( in Figure 2.3 (a)) of a square

channel, and the Capillary number using an asymptotic method based on experimental [41], [42]

and computational studies [43].

(a) (b)

Figure 2.3 Cross-sectional view of a bubble in a square microchannel (a) Non- axisymmetric bubble (b) Axisymmetric bubble (Kolb and Cerro, 1991 [41]).

13 , 0.7 0.52.25. (2.12)

In gas-liquid two-phase flow with Capillary numbers less than 0.04, Hazel and Heil [43]

calculated the scale factor numerically, ⁄ 0.99 for the diameter of the bubble in

the direction in Figure 2.3 (a) and Kolb and Cerro [41] determined the scale factor

experimentally, ⁄ 0.95.

Figure 2.4 (a) shows the dimensionless film thickness as a function of the Capillary

number in circular and square (direction ) channels as proposed by Bretherton (1961),

Aussilous and Quére (2000), and Hazel and Heil (2002). Figure 2.4 (b) shows the dimensionless

film thickness as a function of the capillary number in a square capillary in directions and

in Figure 2.3 (a). The length of the bubbles is much larger than channel width in the slug

flow regime. Kolb and Cerro [41] experimentally studied the elongated bubbles in square

capillaries through the direction of and in Figure 2.3 (a). They also showed that the transition of bubble shape from non- axisymmetric (Figure 2.3 (a)) to axisymmetric bubbles

(Figure 2.3 (b)) occurred at a Capillary number of around 0.1.

0.01 0.1 1 1 1 Aussilous and Quere (2000) Bretherton (1961) Hazel and Heil, bb' (2002)

0.1 0.1 d δ/

0.01 0.01

0.01 0.1 1 Capillary number, Ca (a) (b)

Figure 2.4 Scaled liquid film thickness between gas bubble and channel walls in (a) a capillary tube and (b) square channel in function of capillary number (Ca) (Kreutzer et al., 2005 [39]).

14 2.5 Literature Review

This section provides a broad review of previous work from fundamental studies to the practical applications of the gas-liquid two-phase flow and liquid-liquid segmented flow in bioanalytical systems.

2.5.1 Gas-Liquid Two-Phase Flow

Günther et al. [21, 22] used micro particle image velocimetry (µPIV) to characterize gas- liquid two-phase flow in PDMS/glass microchannels and verified Taylor’s prediction [36] of the recirculation of streamlines and stagnation points in liquid plugs segmented by gas bubbles [see

Figure 2.5]. They also reported a gas-liquid two-phase flow regime map.

Visualization and measurement of gas-liquid two-phase flow patterns for nitrogen and

15 four different liquids (deionized water, ethanol, 10% and 20% glycerol) in rectangular silicon

microchannels with hydraulic diameters raging from 187.5 to 218 µm and aspect ratios up to 2.7

was done by Wälchli and Rohr [17]. Their study confirmed the conclusions of Zhao and Bi [44]

that the intermittent regime transition to Annular flow was shifted to higher gas superficial

velocities with decreasing hydraulic diameter. They developed a flow pattern map based on

dimensionless numbers and reported good agreement for all of the fluids and hydraulic diameters.

Liquid velocity distributions inside the liquid plugs were measured using µPIV and the

recirculation inside the liquid plugs was affected more by interfacial forces than by the viscosity

of the liquid.

Cubaud and Ho [19] investigated gas-liquid two-phase flows in 200 and 525 µm square microchannels fabricated in silicon and glass. A two-phase flow map, identifying transitions between flow regimes, was defined. They also measured and discussed bubble velocity, volume fraction, slip ratio, and pressure drop characteristics as well as local and global drying of the partially-wetting channel walls. Zhao and Bi [44] characterized upward air-water two-phase flow patterns in three vertical, equilateral triangular cross-section channels with hydraulic diameters of 2.89, 1.44 and 0.87 mm. While typical flow patterns such as dispersed bubbly flow, slug flow, churn flow, and annular flow were found in the larger 1.44 and 2.89 mm hydraulic diameter vertical triangular channels, no dispersed bubbly flow was observed in the smallest channel.

Instead of dispersed bubbly flow, more regularly distributed bubbles along the channel centreline in the liquid flow, which was called capillary bubbly flow, were found. With the decrease in the hydraulic diameter of the vertical triangular channels, transition boundaries from slug to churn and from churn to annular flow regimes were shifted to higher gas and liquid superficial velocities on the flow regime map. Effects of diameter and cross-sectional geometry of channels on the transition of air-water flow in horizontal capillaries ranging in hydraulic diameter from

16 5.5 to 1.3 mm were investigated by Coleman and Garimella [45]. While previous studies [46, 47]

concluded that the effects of pipe diameter and surface tension were negligible for pipes with

diameters larger than 10 mm, their study showed that diameter and surface tension effects were principal factors in determining flow patterns and transition in smaller tubes less than 10 mm

diameter and transitions in these flow patterns occurred at different combination of superficial

gas and liquid velocities with decreasing of tube diameter.

Kawahara et al. [24] carried out experimental investigations on nitrogen and de-ionized water two-phase flow in 100 µm diameter capillary tubes made of fused silica. They observed mainly intermittent and semi-annular flows in the capillary tubes. With a closer look at the liquid film between the gas core and tube, they identified additional flow patterns such as gas core flows with a smooth or ring-shaped film and a serpentine-like gas core flow surrounded by a deformed liquid film. No bubbly flow was reported. They also compared the single-phase frictional pressure drop factor and two-phase friction multiplier data with conventional correlations. Their results for single-phase friction factor showed good agreement with conventional laminar correlations. Two-phase friction multiplier data correlated well with the separated flow model of Lockhart and Martinelli [48] who postulated a classic correlation of

two-phase flow in pipes. Lee and Lee [49] proposed new correlations of two-phase (air-water)

frictional pressure drop in horizontal rectangular channels with gaps between the upper and

lower plates ranging from 0.4mm to 4 mm for a 20 mm fixed channel width. These test sections

ranged in hydraulic diameter from 6.67 mm to 0.78 mm. The correlations of frictional pressure drop of the gas-liquid flow were Lockhart-Martinelli type with a newly defined parameter C. The correlation represented the gas-liquid frictional pressure drop for the Matinelli parameter (X) over Reynolds number ranges of 0.303-79.4 and 175-17,700, respectively, in the narrow channels where the surface tension effects became an important factor.

17 Table 2.1 Gas-liquid two-phase flow test parameters in literatures

Gas Liquid Cross Hydraulic Source Dimension Material Gas-Liquid superficial Superficial section diameter velocity (m/s) velocity (m/s) 5.5, 2.6, 1.75 and Coleman and 1.3 mm diameters, Circular, Glass and Garimella hydraulic diameter 1.3 – 5.5 mm Air-water 0.1-100 0.01-10 Rectangular polymer (1999) of 5.36 mm rectangular

20 mm in width Lee and Lee Acrylic Rectangular and gags between 0.78-6.67 mm Air-water 0.05-18.7 0.03-2.39 (2001) plate 0.4-4 mm

Zhao and Bi Equilateral 1.5, 2.5 and 5 mm 0.866, 1.443 Lucite® Air-water 0.1-100 0.08-10 (2001) triangular side length and 2.886 mm plate

Kawahara et al. Circular 100 µm diameter 100 µm Fused silica Nitrogen-water 0.1-60 0.02-4 (2002)

Cubaud and Ho 200 and 525 µm 200 and 525 Square Silicon Air-water 0-100 ml/min 0-1 ml/min (2004) square µm

Nitrogen – de- 150 µm in depth Waelchli and Rohr ionized water, Rectangular and widths between 187.5 – 218 µm Silicon 1-20 ml/min 0.05-5ml/min (2006) ethanol and 150-400 µm glycerol

18 Table 2.1 summarizes test parameters for the experimental investigations of gas-liquid

two-phase flow in microchannel in the literature.

2.5.2 Liquid-Liquid Segmented Flow

Thorsen et al. [50] studied experimentally the formation of emulsion by shearing one

liquid (water) into another immiscible liquid, a mixture of oil and surfactant, through a T- junction. Decane, tetradecane and hexadecane mixed with the surfactant (Span 80) in concentrations of 0.5%, 1.0% and 2% (v/v) were tested as immiscible liquids. The size of the water vesicle and generating frequency was controlled by changing the relative pressure of each water and oil/surfactant stream in 35 µm width and 6.5 µm depth microchannels fabricated in acrylated urethane. The distribution and morphology of the droplet pattern was also affected by channel shape. Song et al. [51] demonstrated methods for controlling the merging of two aqueous plugs into a plug in a main stream by using relative flow rates of different size of plugs

[see Figure 2.6 (a)]. Since small droplets flow more slowly than large droplets [50] in a stream of carrier fluid (PFD, perfluorodecalin), two different sizes of aqueous droplets were coalesced to make a plug. Additionally, they studied splitting of plugs into two smaller plugs by constricting the channel at a T-shaped branching point [see Figure 2.6 (b)]. The size of the split plugs depended on the relative flow rates (or pressure) in the two outlet channels. Link et al. [23] presented an analytical model for the passive splitting of plugs at a T-shaped junction and demonstrated asymmetric splitting of droplets by adjusting the relative lengths of the two side channel outlets or by controlling the relative pressure gradients in the two channels.

There have been many attempts to replace conventional 96-well micro-titer plates with the continuous segmented flow in microchannel for high throughput screening in addition to the reduced time, reagent volumes and human involvement. Chen and Ismagilov [10] reported a

preloaded microfluidic cartridge [see Figure 2.7]. The aspirated nanoliter plugs of reagent in a capillary tube cartridge were flowed into a T-shaped merging junction. Each reagent was separated by fluorinated oils to prevent cross-contamination between the adjacent reagent plugs.

The T-shaped merging junction was used to combine the separate plugs and another substrate solution in a 1:1 volumetric ratio. Using the serially preloaded reagent cartridge, they tested a fluorescence assay of enzymes and screening conditions for protein crystallization. Li et al. [52] from Dr. Ismagilov’s research group also developed a process for screening of membrane proteins through crystallization, which were confined in plugs with different concentration of a reagent. Figure 2.8 shows the different patterns of precipitation of protein for different concentration of reagent in each plug.

Figure 2.8 Screens of crystallization conditions for membrane protein with different concentration of precipitant in each plug, which showed different crystal patterns, (Li et al., 2006 [52]).

21 In efforts to automate the reagent delivery in droplet microfluidics, Linder et al. [11] have

developed “plug-in cartridge” technology for storing and delivering a series of reagent plugs to a microfluidic device. They injected a sequence of reagent and air spacers, which prevented the

reagents from mixing with each other, into a 30 cm long polyethylene (PE) tube using a manual

syringe or an array of valves connected to a vacuum as shown in Figure 2.9 (a). After connecting

the cartridge to a microfluidic chip, vacuum was applied to an outlet so that the sequence of

reagent and air spacer could be introduced into the microchannel smoothly. They also injected

three washing buffer plugs between the reagents to minimize the cross-contamination between

the adjacent reagent plugs. Alternatively, perfluorodecalin (PFD, a water-immiscible liquid) was

used instead of air as a spacer between reagent plugs. However, PFD plug didn’t flush completely the aqueous reagent left on the microchannel. Air spacers had another advantage since they traveled through the microchannel ~250 times faster than PFD under the same pressure gradient because of the viscosity difference. Using this “plug-in cartridge” method, they performed an immunoassay for Anti-Rabbit IgG and detected a clear fluorescence signal from the area of the immobilized antigen (rabbit IgG) positioned perpendicular to a microchannel and no signal outside the area. Adamson et al. [12] produced arrays of nanoliter plugs using several

T-junctions in a microfluidic chip [see Figure 2.9 (b)], which could be used for high throughput

biological screening. A single input array of large plugs in distinct chemical compounds was

split passively into 16 output arrays of smaller plugs whose chemical composition and

configuration were identical to the input array plugs. In order to prevent coalescence of adjacent

plugs as a result of different viscosities causing relative flow rates between two consecutive

plugs, gas spacers were introduced between aqueous sample plugs. While the gas spacer

prevented adjacent plugs from coalescing, there was still cross-contamination between plugs.

(a) (b)

With a higher Capillary number (Ca), small droplets were detached from plugs and stuck to the microchannel surface. Even the gas spacer could not flush out the droplets from the channel wall and there was mixing with following plugs. To avoid the breakup of plugs resulting in cross- contamination, a low Capillary number should be maintained.

Curcio and Roeraade [9] utilized continuous segmented flow in the polymerase chain reaction (PCR) for high throughput genetic analysis. They loaded separated liquid samples in the

15 m long Teflon tube circulating 30 times within three static temperature zones intended to reproduce the PCR thermal-cycling process. Reaction mixtures were loaded manually using compressed helium gas and two consecutive samples were separated by hydrocarbon (n-decane) or fluorocarbon (perfluorodecalin). Water plugs were also injected between samples to minimize

23 the cross-contamination by minute droplets detached from preceding sample plugs. Results of

DNA amplification showed that perfluorodecalin was preferred to reduce the cross- contamination because fluorocarbons are more hydrophobic than hydrocarbons. The hydrophobic property increased the surface tension between the carrier fluid and aqueous sample, which restrained aqueous sample from disintegration. Even though cross-contamination was significantly reduced by using perfluorodecalin as a carrier fluid, water plugs between the aqueous samples were still required to get acceptable amplification results without cross- contamination. Mohr et al. [53], Beer et al. [54], Gonzalez et al. [55], and Kiss et al. [56], have adapted droplet flow to develop chip-based PCR in purpose of high throughput analysis.

Utada et al. [57] generated a double emulsion by aligning a circular injection and collection tube in a square outer tube. The microcapillary device generated larger droplets encapsulating smaller droplets. The size of the inner and outer droplets, as well as the number of droplets encapsulated in each larger drop, could be controlled by modifying flow rates of the inner and middle fluid. Nisisako et al. (2006) [58] also generated double emulsions in order of water-in-oil-in-water and oil-in-water-in-oil by switching hydrophobic and hydrophilic T- junctions. Generation of aqueous droplets in an organic stream, a corn oil, in hydrophobic T- junction to prevent aqueous droplets from wetting the channel surface and being carried to the hydrophilic T-junction where organic droplets encapsulating smaller aqueous droplets were generated. Control of flow conditions enabled two aqueous droplets to be encapsulated within an organic droplet.

Table 2.2 summarizes experimental parameters and applications of liquid-liquid segmented flow in microchannels shown in the previous literature.

24 Table 2.2 Test parameters and applications of liquid-liquid segmented flows.

Cross Sample Source Research group Dimension Material Carrier fluid Application section phases 35 µm in width n-decane with Thorsen et al. Acrylated Water vesicle S. R. Quake’s Rectangular and 6.5 µm in Water surfactant (Span (2001) urethane generation height 80) Curcio and Continuous 200 µm Roeraade J Roeraade’s Circular Teflon® Tube PCR cocktail Perfluorodecalin segmented- diameter (2003) flow PCR 25 µm and 100 Perfluorodecalin Merging and Song et al. R. F. Ismagilov’s Rectangular in width and 45 PDMS Water with 9% (v/v) of Splitting of (2003) µm in height C6F11C2H4OH droplets Hexadecane with Passive Link et al. Square 30 µm square PDMS Water surfactant (Span splitting of (2004) 80) droplets D. A. Weitz’s 1mm square, Silicon oil in Utada et al. Square and Double 10-500 µm Glass glycerol-water Silicon oil (2005) circular emulsions diameter mixture Nisisako et al. Water in corn Double T. Torii’s Square 40-200 µm Glass Water (2005) oil emulsion Delai and Cartridge 200 µm Aqueous Ismagilov Circular Teflon® Tube Perfluorodecalin loaded with diameter reagents (2006) nanoliter plugs 200 µm R. F. Ismagilov’s Production of diameter, 250 Mixture of Adamson et al. Circular and Teflon® Tube Aqueous arrays of µm in height perfluorooctanol (2006) rectangular and PDMS reagents nanoliter plugs and 50-1000 and FC 3283 by splitting µm in width

25 Chapter 3 Polymer Microfluidic Test Chips and Experimental Apparatus

3.1 Introduction

Modern technology provides several ways to fabricate microfluidic chips from various

materials [25, 59, 60]. Etching of silicon and glass [15, 19, 61, 62], soft lithography using

polydimethyllsiloxane (PDMS) [63-67], laser ablation of polymers [68, 69] or metal molds for

thermoforming [70, 71], direct micromachining of polymers using high-precision computerized

numerical control (CNC) machine [72], micromachining of metallic mold inserts for

thermoforming of polymers [14, 73], ultraviolet (UV) [74] or X-ray LIGA processing of

polymers [75-77] are currently favored methods for microfluidic chip fabrication. Polymeric

microfluidic chips were fabricated by three different methods; 1) Direct micromachining of

polymers; 2) Micromachining of brass mold inserts followed by hot-embossing; and 3) X-ray

LIGA process. These different fabrication methods allowed for comparison of chip

characteristics in terms of surface qualities of the microchannels, inherent limitations caused

from present fabrication equipment, process time, and cost. Each fabrication process is presented

in this chapter.

Polymer microfluidic chips were fabricated for the experimental investigations of

multiphase flows in microchannels. For this purpose, highly controllable and sensible injection

systems for the steady and stable fluid supply, transducers with high sensitivity and resolutions

for the pressure drop measurements, and microscopic optical systems for the image acquisition

and processing were essential to get reliable experimental data. Details of each experimental

apparatus for the gas-liquid two-phase flows and liquid-liquid segmented flows in polymer

microchannels were provided in this chapter.

26 3.2 Preparation of Polymer Microfluidic Test Chips

3.2.1 Direct Micromachining of Polymers

The fastest way from a design to a polymer microfluidic chip is to engrave channels on

the polymer directly using computerized numerical control (CNC) milling machine or laser

ablation. Computer-aided manufacturing (CAM) software converts original CAD design files

into the G-code which drives machine tools. The G-code is interpreted by controllers of CNC

milling machines and organize machining steps under some restrictions like limited drill length,

rotation speed, feed rate and the radius of drill bit resulting in curvatures at the concave edge of

structures [78, 79]. Microchannels were directly machined in the polymer chip using the micro-

milling and drilling machine (KERN MMP 2522, KERN Micro- und Feinwerktechnik GmbH &

Co. KG, Murnau-Westried, Germany) shown in Figure 3.1. Even under the same machining

parameters such as spindle speed, radius of drill bit and feed rate, the surface qualities of

(a) (b)

(a) (b) (c)

Figure 3.2 (a) Micromachined polycarbonate chips. (b), (c) Scanning electron microscope (SEM) images of Burrs along the 200um channels generated from machining on the polycarbonate.

channels depend on the materials to be machined. In this fabrication process, polycarbonate (PC,

LEXAN®, GE Plastics, Pittsfield, MA) and poly(methyl methacrylate) (PMMA, ACRYLITE®

FF acrylic sheet, Cyro Industries, Rockaway, NJ ) were used as polymer microfluidic chip

materials. Many applications in polymer microfluidic chips are based on PC [75, 80, 81] and

PMMA [82-84]. The physiochemical properties of polymers, including PC and PMMA, have

been reported by Shadpour et al. [85]. Both of these polymers are amorphous, dimensionally

stable, apparently transparent, and easily thermoformed. In spite of these physical similarities

between PC and PMMA, different surface qualities of microchannels were observed after direct

micromachining. Figure 3.2 shows (a) a photograph of directly machined PC chips and (b) SEM

(scanning electron microscopy, S-3600N, Hitachi, Pleasanton, CA) images of burrs which are

generated from the milling process along channels. These burrs were hard to remove from the

channels using ultrasonic agitation, which usually showed a good cleaning result by removing

burrs in polymer chips. On the other hand, PMMA generated much fewer burrs during

micromachining and it was relatively easy to remove the burrs from the channels through

ultrasonic agitation. Clear channels after ultrasonic agitation were observed through SEM images

(a) (b) (c)

in the Figure 3.3. These different surface qualities as a result of micromachining can be

explained in terms of the elastic properties and monomer chains of the polymers. PMMA is much stiffer, more brittle and less elastic than PC. Less elastic polymer materials which have shorter monomer chains compared to more elastic ones easily break into small shards without stickiness when external forces are applied [86]. These shards of polymer can be easily cleaned by ultrasonic agitation. PMMA was chosen for the direct micromachining of the polymer chips because of its better surface quality after micromachining. Milling bits (PMT, Janesville, WI) with certain radii range from 25µm to 200µm resulted in concave curvature at the edge of the microchannels as shown in Figure 3.3 (b). Fabrication of the direct micromachining of polymer chips and fluidic experimental study were carried out by Estelle Evans, who was a collaborator and MS candidate.

3.2.2 Micromachining of the Brass Mold Insert

While the direct micromachining of polymer generates only one polymer microfluidic chip from each machining cycle, fabrication of mold inserts followed by thermoforming enables replication of sufficient polymer chips as long as the machined microstructure maintains its geometric integrity without deformation or wear from the thermoforming process.

(a)

(b) (c)

(d) (e)

30 Instead of polymer substrates, brass plates (Alloy 353, McMaster-Carr, Atlanta, GA), an alloy of

copper and zinc, were machined using the same micro milling and drilling machine described in

the previous section. Brass stock was cut into 4.75 inch diameter disks with six threaded holes

used to mount the brass mold insert in the hot embossing machine. Figure 3.4 (a) shows a typical

brass mold insert with micromachined structures. SEM images of inherent curvature from the

drill bit on the edges of the brass microstructures is shown in Figure 3.4 (b) and Striations produced by rotation of the drill bit are observed in Figure 3.4 (c). These curvatures and striations were replicated in the hot-embossed polymer chip (Figure 3.4 (d) and (e)). Hot embossing of polymer substrate is a stamp-like process generating inverted geometries from the mold to the polymer. This inverted channel shape resulted in the opposite curvature at the edges in the polymer chips. These concave and convex curvatures from the direct machined chip and hot embossed chip with brass mold insert can be compared from Figure 3.3 (a), (b) and Figure

3.4 (d). Striations transferred from mold to polymer (Figure 3.4 (e)) resulted in surface roughness along the microchannels. Hot embossing and surface roughness are discussed in Sections 3.2.4 and 3.3, respectively.

3.2.3 X-Ray LIGA

X-ray LIGA stands for X-ray lithography, Electroplating, and Molding (X-ray

LIthographie, Galvanoformung and Abformung in German) [60]. As pioneers in the areas of X- ray LIGA, E. W. Becker and W. Ehrfeld [87, 88] developed separation nozzle systems using X- ray LIGA in the early 1980s. The high energy of parallel X-ray radiation has some advantages in fabrication of microstructures requiring exact dimensions among structures packed in a limited foot print, high resolution, low surface roughness, and high aspect ratios that are hard to achieve with other microfabrication methods [60, 89, 90]. X-ray exposure of thick photoresist and

32 development of physically modified photoresist gives a patterned master template for metal electroplating. After removing the remained photoresist worked as a master, the electroplated metal structure functions as an application in MEMS [91, 92] or a mold insert for the molding of thermoplastics [93, 94]. In this work, the LIGA process was used to fabricate nickel mold inserts

to replicate polymer microfluidic chips through hot embossing. The molded channel surface was

characterized with scanning electron microscopy (SEM) and optical profiler meters and showed

the smoothest surface in terms of surface roughness compare to other methods for the fabrication

of polymer chips. Surface roughness was similar with etched channel in silicon and glass

substrates in RMS values around 40nm. Schematic overview of the X-ray LIGA process is

shown in Figure 3.5 and details of fabrication steps including X-ray mask fabrication, X-ray

exposure, nickel electroplating and post-processing are provided in the Appendix A.

3.2.4 Hot Embossing of Thermoplastics

Hot embossing is one of the most commonly used thermoforming methods for high

volume production of low cost polymer microfluidic chips. While hot embossing requires more

manual work and longer process cycle times than injection molding, another thermoforming

method used in industrial production of polymer components [95], more precise, delicate and

high aspect ratio microstructures can be replicated as a result of low polymer material flow

during embossing [95, 96]. A 3 mm thick polycarbonate (PC, LEXAN®, GE Plastics, Pittsfield,

MA) and poly(methyl methacrylate) (PMMA, ACRYLITE® FF acrylic sheet, Cyro Industries,

Rockaway, NJ) sheet was thermoformed by using a commercial hot embossing machine (HEX

02, JENOPTIK Mikrotechnik, Jena, Germany) at CAMD shown in Figure 3.6. Micromachined

brass or LIGA mold inserts were mounted underneath the upper plate (Figure 3.6 (b)) and the

polymer sheets were placed on the bottom plate (Figure 3.6 (c)) of the embossing machine.

(b)

(a) (c)

Figure 3.6 Hot embossing machine at CAMD. (a) HEX 02 (JENOPTIK Mikrotechnik, Jena, Germany). (b) Upper plate for mold insert. (c) Bottom plate for polymer to be embossed.

Embossing temperature and force, holding time and demolding temperature are the most important parameters in hot embossing [25]. These parameters vary in some ranges depending on the structural geometry of the microstructure and material properties of polymer substrates and metal mold inserts. Typical values of hot embossing parameters for PC and PMMA are listed in

Table 3.1 [96]. Temperatures of the upper and lower plates were set separately to above the embossing temperature. Following chamber closing, the chamber was evacuated. Once the temperature of the plates reached the embossing temperature, the mold insert was pressed into the polymer with an embossing force for a specific holding time. After both plates were cooled down and reached demolding temperature, the mold insert was demolded. Opening of the

chamber completed a cycle of the hot embossing process. Hot embossing using the HEX 02 at

CAMD required different values of the parameters in Table 3.1. The temperature sensor just

measured the temperature of the plate, not of the polymer substrate, thus there were some

34 Table 3.1 Properties and hot embossing parameters for PMMA and PC [96].

Polymer PMMA PC

Specific Density 1.17 – 1.20 1.20

Glass transition temperature (°C) 106 150

Young’s modulus (MPa) 3,100 – 3,300 2,000 – 2,400

Embossing Temperature (°C) 120 – 130 160 – 175

Demolding Temperature (°C) 95 135

Embossing force (4″) (kN) 20 -30

Holding time (s) 30 – 60

differences in the embossing and demolding temperatures. To account for heat loss between

polymer and bottom plate, mold insert and upper plate, temperatures of both plates were set

higher than the temperatures in Table 3.1 [e. g., in case of PC, 185 °C for embossing temperature

and 145 °C for demolding temperature].

3.2.5 Thermal Fusion Bonding and Interconnection

Thermoforming of polymer with a mold insert provided just open microchannels in a

substrate. In order to enclose the microchannels, a thin film of the same material was thermally

bonded without any adhesive. Thermal fusion bonding was usually achieved above the glass

transition temperature (Tg) of the thermoplastic by mobilizing polymer chains [97, 98] and

developing bond strength through interdiffusion and cross-linking of the mobilized polymer chain between intimately contacting surfaces of the polymers [99]. Thermal fusion bonding of polymers represented high dependence on temperature, time and applied pressure. Excess of these parameters during thermal bonding resulted in deformation and collapse of thin films and

Figure 3.7 Thermal bonding jig

microchannels. Sufficient bond strength with minimum deformation through a low value of these

parameters required long bonding times typically more than 24 hours [100]. Since the long

bonding time was not desirable in the fabrication process, bonding conditions minimizing

bonding time and deformation of the polymer test chips were determined empirically by using a

custom-made bonding jig. Figure 3.7 shows the thermal bonding jig with four holes for the

bolting of two plates. A 0.375″ Aluminum plate (Alloy 2024 Aluminum, McMaster-Carr,

Atlanta, GA) was machined to make the thermal bonding jig for maintaining and quantifying the

applied pressure during the bonding process. Before thermal bonding, 1/16″ diameter holes were

drilled in the inlet and outlet chambers of the polymer chips to be connected with macro-scale

external apparatus through capillary tubes. Polymer chips covered by 125 µm thin film was placed between 1/4″ thick borosilicate glass plates (8476K135, McMaster-Carr, Atlanta, GA) providing flat surfaces and intimate contacts between the film and chip. Spring disks between flat washers compensated for any gap which might be produced during thermal bonding. A torque driver with a capacity of 1.5 lb·in with dial increments of 0.03125 lb·in (Precision

36 Instruments, Inc., Des Plaines, IL) was used to apply constant torque in each bonding process.

Each bonding process followed the below steps.

Cleaning

Ultrasonic agitation (Model 2510, Branson Ultrasonics Corp., Danbury, CT) of polymer chip in

the below solutions by turns;

1. Solution of de-ionized water + Detergent (1% volume) (Liqui-Nox®, Alconox, Inc., White

Plains, NY) for 5 minutes to remove organic contaminants.

2. Solution of de-ionized water + Isopropyl alcohol (2:1 in volume) for 3 minutes to remove

detergent.

3. De-ionized water for 3 minutes to remove isopropyl alcohol.

Dry and Dehydration

After remained water was dried out by a gentle stream of nitrogen gas, polymer chip was put in a

convection oven at 95 °C for 10 minutes for dehydration of water molecule

Torque and Preheating

Place a polymer chip enclosed with a thin film in the jig and apply equal torque (0.25 lb·in) on

four bolts using torque driver and put the jig in the convection oven in temperature of 95 °C for

10 minutes to preheat the jig and polymer chip.

Thermal Bonding and Cooling

Set the bonding temperature shown in Table 3.2 for the corresponding polymer. Boding time was

measured after the temperature reached the set point of the thermal controller of convection oven.

Following the complete of required bonding time, temperature was set to room temperature and

the polymer chip was cooled down inside the closed oven. Conical NanoportsTM, shown in

37 Table 3.2 Thermal fusion bonding conditions for PMMA and PC.

Torque Bonding Bonding Material Preheating (lb·in) Temperature (°C) Time (minutes)

PMMA 106 0.25 95 °C for 10 min 100 PC 152

(a) (b)

Figure 3.8 (a) (N-333, Upchurch Scientific, Inc. Oak Harbor, WA), provided interconnections

between the microchannels and conventional tubes (0.03″ inner diameter, 1/16″ outer diameter

(OD) Teflon® PFA and PEEKTM tube, 0.007” (175 µm) inner diameter, 1/32” outer diameter

PPEKTM, Upchurch Scientific, Inc. Oak Harbor, WA). Alternatively, only a Teflon or Peek tube was inserted into the drilled hole and sealed with Devcon 5 minute® Epoxy (Figure 3.8 (b)).

3.3 Surface Roughness of Microchannels

Each polymer test chip showed different surface roughness as a result of the different

fabrication methods. Surface roughness parameters, average roughness (Ra) and average root

mean square (RMS), of the sidewall and bottom of the microchannels were measured by using an

38 optical profiler, Wyko NT 3300 (Veeco Instruments Inc., Woodbury, NY), at CAMD. Average

roughness (Ra) is an average height of the bumps on a surface calculated over the entire

measured area (Equation 3.1). Root mean square (Rq, rms) average roughness was the square

root of the mean of the squares of the bump heights on a surface (Equation 3.2).

1 | | (3.1)

1 (3.2) where n is number of data points and Z is the surface height relative to the mean plane ( Z ).

Figure 3.9 and 3.10 shows the scanning electron microscope (SEM) and optical profiler images for the bottom of the chamber and the sidewalls of the microchannels in each polymer microfluidic chip. Figure 3.11 shows roughness profiles of the bottom and sidewall of a microchannel along the center line of the x-direction shown in (d), (e) and (f) of the Figure 3.9 and 3.10. Surfaces of the etched silicon microchannels had the root mean square (RMS) roughnesses varying between 5 and 50 nm [101], which is close to a molecularly smooth surface and a difference from a polymer chip. Specific average and RMS values for the sidewalls and the bottoms of the channels in each polymer chip are shown in Table 3.3. These parameters were

acquired from 353,280 data points (736 × 480) over an area of 119.881 µm × 91.174 µm with

163.1 nm × 190.34 nm sampling distance through a 51.50X objective. In order to make the

polymer surface reflective to the light source of the optical profiler, an approximately 20 nm thick gold layer was coated on the surface of the polymers using a sputter coater. Surface roughnesses for the sidewalls of the microchannels in the direct micromachined chip and chip hot embossed with a micromachined brass mold insert were introduced by striations generated

(a) (b) (c)

(d) (e) (f)

(a) (b) (c)

(d) (e) (f)

40 with the cutting tool during the machining process. On the other hand, sidewalls of the microchannels in the hot embossed chip with the X-ray LIGA mold insert were very smooth and close to the value of microchannels etched in silicon. The smooth sidewalls are one of the characteristics of the X-ray LIGA process. Surface roughness values for bottoms of chambers in the hot embossed chips were mainly subject to the top surface roughness of the prepared substrates for fabrication process. The top surface roughness of the substrate was exactly reflected on the bottom of the channel through the hot embossing process.

Table 3.3 Surface roughness values of polymer chips.

Sidewall of channel Bottom of channel

RA (nm) RMS (nm) RA (nm) RMS (nm)

Direct micromachined chip 309.14 384.42 241.36 302.46 Hot embossed chip with a 331.61 402.62 147.16 208.92 micromachined brass mold insert Hot embossed chip with LIGA 63.01 78.77 539.99 688.36 nickel mold insert

(a) (b)

41 3.4 Experimental Apparatus

3.4.1 Gas-Liquid Experiment Setup

A schematic of the experimental setup used in the gas-liquid two-phase flow is shown in

Figure 3.12. Deionised water was used as the working liquid and driven from a thoroughly

cleaned PVC liquid reservoir. Flow rate of liquid from the liquid reservoir was subjected to the

pressure applied by building air supply. Compressed air through a filter (Finite® filter 11F12EC,

Parker Hannifin Corp, Cleveland, OH) to reservoir was controller by a pressure regulator (14R

Manual, 0-150 psig, Parker Hannifin Corp, Cleveland, OH). The liquid flow rate, QL, is measured by a 0-1 mL/min flow meter (maximum operating pressure of 100 psig ± 1% full scale accuracy, and 100 msec response time; 32908-40, Cole-Parmer, Vernon Hills, IL) positioned between the liquid reservoir and the microchip inlet port. In order to guarantee safe operation of all the instrumentation and the microchip itself, the maximum supply pressure was limited to 90 psig.

Figure 3.12 Schematic of the experimental apparatus for gas-liquid two-phase flow

42 A pressurized liquid reservoir was preferred (as in Cubaud and Ho [19]) to a syringe

pump because the latter displayed oscillatory behaviour (modulated liquid flow-rates), which can

influence the two-phase flow generation. Furthermore, the syringe pump provides a limited flow

time because of syringe size limitations and steady state was obtained slower than when the

pressurized reservoir was used. A pressure-regulated dry air cylinder (Water < 6ppm; Matheson

Tri-Gas, Inc., Houston, TX) was used as the gas source. The air was distributed to the gas inlet port of the microchip by one of two flow controllers (maximum operating pressure of 100 psig, ±

1% full scale accuracy, and 100msec response time; 32907-51 and -59, Cole-Parmer, Vernon

Hills, IL) connected in parallel. A three-way solenoid valve (Model 155, Peter Paul Electronics

Co. Inc., New Britain, CT) directed the flow through the appropriate flow controller. The two flow controllers covered the complementary ranges of 0-1mL/min and 1-100mL/min and thus provided a wide range of adjustable gas flow. Operating pressure and pressure drop in the microchannel was measured using gage and differential (0-5 psi and 0-15 psi) pressure transducers (ASDX Series, Honeywell Inc., Morristown, NJ) respectively.

Table 3.4 Description of objectives performance

Numerical Working Cover Glass Objective Magnification Aperture Distance Thickness Immersion (N.A.) (W.D., mm) (mm) PlanApoN 2X 0.08 6.2 - -

UPlanSApo 4X 0.16 13 - -

UPlanFl 10X 0.30 10 - -

LUCPlanFLN 20X 0.45 6.6-7.8 0 - 2 -

UPlanFl 40X 0.75 0.51 0.17 -

PlanApo 60X 1.40 0.15 0.17 Oil

43 The experiment is computer-controlled as is the data acquisition from all the instrumentation through a digital acquisition and I/O system (FieldPoint system, and I/O board,

National Instruments Co., Austin, TX), with the aid of LabVIEW® Virtual Instrument software modules developed in-house.

The observations and measurements reported here were conducted on an inverted fluorescence microscope (IX70, Olympus, Ceter Valley, PA) using 2X, 4X and 10X objectives and equipped with a digital camera (SharpVISION 1400-DE, Integrated Design Tools, Inc.,

Tallahassee, FL). Performances of objectives used in this study were listed in the Table 3.4.

Continuous back-illumination (Hg-lamp or fiber-optic light-guide with diode array and diffuser) is used to acquire flow images for topological observations and bubble and liquid plug dimension statistics. Quantitative information was extracted from the acquired images by using image processing routines developed in house and based on OPTIMASTM (ver. 6.51, Media

Cybernetics, Inc., Bethesda, MD) image processing software.

3.4.2 Liquid-Liquid Experiment Setup

The experimental setup for the liquid-liquid segmented flow is shown schematically in

Figure 3.13. Carrier and dispersed phase fluids were driven into test chip from glass syringes

(Gastight® 1000 Series, Hamilton, Reno, NV) through capillary tube using syringe pumps

(NE500, New Era Pump Systems, Inc., Wantagh, NY). The pressure drop was measured using differential pressure transducers (0-5 psid and 0-15 psid, ASDX Series, Honeywell Inc.,

Morristown, NJ) connected to the up and downstream pressure taps fabricated on the test chips.

The values of pressure drops were recorded through a data acquisition systems (FieldPoint and

I/O board, NI Co, Austin, TX) with aid of LabVIEW Virtual Instrument software modules developed in-house. The optical observations and measurements reported here were conducted

Figure 3.13 Schematic of the experimental apparatus for liquid-liquid segmented flow.

on inverted fluorescence microscopes (IX70 and IX81, Olympus, Center Valley, PA) using 2X,

4X and 10X objectives and equipped with a digital CCD cameras (SharpVISION 1400-DE and

Rolera-MGi, QImaging, Surrey, BC, Canada). Different illumination sources were used for the

specific observation purposes; 1) Continuous white field back-illumination (fiber-optic light-

guide with diode array and diffuser) for the acquisition of clear flow images for topological

observations and droplet and plug dimension statistics, 2) double pulsed 532nm Nd:YAG laser

for the measurement of dispersed fluid droplet and plug velocities, and 3) mercury (Hg) lamp for the observation of carrier fluid thin film between dispersed fluid and channel walls by detecting the organic quantum dots (Qdot® 800 ITKTM organic quantum dots, emission: 800nm, Invitrogen

Corp., Carlsbad, CA) suspended in the carrier fluid. For the better optical comparison between

the dispersed and carrier fluids, two different fluorophors were used. 0.2 µm diameter

microspheres (FluoSpheres®, Excitation: 540 nm, Emission: 560 nm, Invitrogen Corp, Eugene,

45 OR) were used for the aqueous dispersed fluid and quantum dots (Qdot® 800 ITKTM organic quantum dots, emission: 800nm, Invitrogen Corp, Eugene, OR) were used for fluorocarbon carrier fluids. Specifications of fluorescence mirror units which were used for corresponding

fluorophors were listed in Table 3.5. Quantitative information was extracted from the images

acquired with white field illumination using the same way as in the gas-liquid experiments using

image processing routines developed in house.

Table 3.5 Specifications of fluorescence mirror units

Excitation Emission Dichromatic Mirror unit Excitation filter filter Mirror

U-MWIG2 Green 520-550 580IF* 565

U-MWG2 Green 510-550 590 570

U-MWU2 Ultraviolet 330-385 420 400

DAPI Ultraviolet 350/50x 460/50m 400

*High-performace interference type filter. It has a steep cutting edge, and excellent spitting characteristics.

46 Chapter 4 Gas-Liquid Two-Phase Flows in Microchannels

4.1 Introduction

Gas-liquid two-phase flow in microfluidic systems is of significant interest due to its

many advantages over single phase flow for enhancing the performance of miniaturized biochemical reaction systems. For better applications of two-phase flows in microfluidic systems, understanding of the fundamental aspects of gas-liquid two-phase flow through experimental study is necessary. The objectives of this experimental study were to characterize gas-liquid two- phase flow in microchannels made of polymer, poly-methyl-methacrylate (PMMA) with walls that were poorly wetting (typical contact angle ≈ 65°C) and not molecularly smooth. The discussion will focus on flow patterns and regimes, with particular interest in segmented flows and their structure and regularity. Pressure drops in single phase and two-phase flow were measured for each flow pattern. The experimental results from the gas-liquid flow in the polymer microchannel were compared to those in the literature using different test materials, dimensions,

and cross-sectional geometries.

4.1.1 Configurations of Microfluidic Test Chips

The configuration of the microfluidic test chip is shown schematically in Figure 4.1 and

the dimensions of the chip are listed in Table 4.1. The configuration of the chip was modified

from that used by Cubaud and Ho [19]. Additional pressure drop channels as used by Günther et

al. [102] were added to the chip to stabilize the generation of two-phase flow. The liquid and gas

injection channels were 50 µm wide by 200 µm deep (±10% considering measured manufacturing tolerances and variability) and the test channels were 200 µm wide by 200 µm

deep, an aspect ratio (AR) of 1. The liquid flows impinging from the both sides meet in a cross junction with the gas flow in the central branch. The two-phase flow was generated in the central

(a)

(b)

48 Table 4.1 Dimensions of the injection and test channels and the surface roughness of the side and bottom walls of the microchannels in the micro-milled and hot embossed chips.

Micro-milled chips [103] Hot embossed chip AR* = 1 AR* = 2 AR* = 3 AR* = 1

Width, w (µm) 50±2.5 50±2.5 50±2.5 50±2.5 Injection i channel Depth, Di (µm) 50±2.5 50±2.5 50±2.5 200±10

Width, w (µm) 200±10 150±6 127±7.6 200±10 Test t channel Depth, Dt (µm) 200±10 300±18 381±22.9 200±10

Nominal hydraulic diameter 200 200 190.5 200 of the test channel, Dh (µm) Distance between - - - 124 pressure ports, ΔZ (mm)

Hot-embossing with micro- Fabrication method Direct micro-milling milled brass mold insert

Side wall RMS roughness ≈ 384 ≈ 403

Bottom wall RMS roughness ≈ 303 ≈ 209

* Aspect ratio of test channels.

branch after the cross junction, which is the same size as the injection channels. This feeds into the test section channel which is arranged in a serpentine configuration wider, and the same depth. Two 50 µm square channels for pressure taps that enabled pressure drop measurements are located on either end of the serpentine test channel. The total length between the two pressure taps is ~124mm. The serpentine configuration was chosen because it is widely used in microfluidic chips due to its compact footprint and since it allows simultaneous observation of the flow over almost the full channel length. The U- and 90o- bends in the path of the test

49 channel have nominal centreline radii of 105µm (+/-6%) which are comparable to the channel hydraulic diameters.

4.2 Gas-Liquid Two-Phase Flow Regimes

Several typical flow regimes have been identified on the basis of observed flow topologies in previous studies of gas-liquid flows in microchannels. These regimes may be summarized in four major categories: Bubbly (capillary or dispersed), Segmented (includes

Wedge, Slug and Plug as used by others, e.g. Cubaud and Ho [19], Günther et aal. [21]), Annular

(includes Wavy-Annular), and Dry. A photographic overview of the observed topologies and their association with specific regime names are given in this section and the field of view extends from the inlet 90o bend to the outlet 90o bend. Vertically it covers the full extent of the straight parts of the serpentine channel.

4.2.1 Capillary Bubble Flow

Figure 4.2 Capillary bubbly flow (CB), [Lb/wt < 1]: gas superficial velocity (JG) ≈ 0.014 m/s, liquid superficial velocity (JL) ≈ 0.093 m/s and the liquid volumetric flow ratio (βL) ≈ 0.87.

50 Flows with spherical bubbles with lengths (diameters) less than the channel width (Lb/wt

< 1) were classified as Bubbly flow. This implied no geometrically-imposed contact of the bubble surface with the wall of the channel. Bubbles in this regime had nearly homogeneous bubble sizes as in Figure 4.2 can be orderly “capillary” flows. Randomly scattered bubble distributions caused by the breakup of the gas flow due to turbulence [44, 45] were not observed due to the laminar flow characteristics in the microchannels.

4.2.2 Segmented Flow

With increasing gas flow rate, the diameter of the bubbles became larger than the channel width (Lb/wt > 1) and the elongated bubbles segmented the liquid flow along the channel. For the

Segmented flow regime, three sub-regimes (-1, -2, and -3) were identified based on the

relationship between the topological length scales and correlations between the three sub-

regimes and the gas-liquid two-phase pressure drops as discussed in the Section 4.5. For the

Segmented-1 sub-regime illustrated in Figure 4.3 (a), the length segment of the bubble which

(a) (b) (c)

51 corresponds to a liquid or corner film is less than the length of the liquid plug that separates the bubbles, combined with the sections of the bubble length which are not part of a film. This was approximated by the ratio (Lb-wt)/(Lp+wt) < 1. In addition the bubble length does not exceed the channel width by an order of magnitude (say Lb/wt < 5). In Segmented-2 [see Figure 4.3 (b)], the ratio (Lb-wt)/(Lp+wt) was larger than unity, but the bubbles were still “short” (Lb/wt < 5).

Segmented-3, as portrayed in Figure 4.3 (c), was associated with “long” bubbles (Lb/wt > 5), which were also large enough to make full contact with all four walls of the channel, satisfying the static condition Lb/wt > 2AR/(1+cosθ). The latter, or one appropriately derived for dynamic conditions, may be relevant for channels with high aspect ratio, where is a distinct possibility that a bubble may only make “contact” with two or three walls. Segmented-3 roughly corresponds to the Slug regime of Cubaud and Ho [19], which was defined using the serpentine channel branch length as the discriminating scale. The Segmented-3 regime evolves into the

Annular regime as the length of the bubbles increases and the segmented liquid plugs become more sparse.

4.2.3 Segmented-Annular Flow

Figure 4.4 Segmented-Annular flow (SA): gas superficial velocity (JG) ≈ 1.72 m/s, liquid superficial velocity (JL) ≈ 0.084 m/s and the liquid volumetric flow ratio (βL) ≈ 0.047.

52 A transition topology denoted as Segmented/Annular flow shown in Figure 4.4, appears as a part of the evolution, in which intermittently occurring, liquid plugs separate segments of essentially Annular flow. The length of the bubbles was much larger than the channel width in

Segmented/Annular flow. Some of the bubble noses began to contact the tails of preceding bubbles. This regime is a transition from Segmented to Annular flow regimes and overlapped with the slug flow identified by Cubaud and Ho [19] based on pressure drop measurements. They distinguished slug from plug flows (wedging flow), when the pressure gradient began to drop abruptly as the liquid plugs were segmented by gas bubbles. Onset of Segmented/Annular flow was defined by the initial observation of Annular flow, which occurred when the liquid plug was initially broken up byy gas bubbles.

4.2.4 Annular Flow

Although the liquid flow rate was constant, increasing the gas flow rate generated longer bubbles with relatively shorter liquid slugs. Elongated bubbles eventually merged with adjacent

Figure 4.5 Annular flow (A): gas superficial velocity (JG) ≈ 4.19 m/s, liquid superficial velocity (JL) ≈ 0.057 m/s and the liquid volumetric flow ratio (βL) ≈ 0.014.

53 bubbles and flowed in a pattern of a continuous gas core surrounded by liquid [see Figure 4.5]. A ring-shaped liquid film appeared where the nose and tail of the bubble met, grew, and disappeared randomly [19]. Churn, wavy, and mist annular (entrained small droplets in a gas core) flows observed between slug and annular flows in relatively large channels (> 1 mm) were not detected in the microchannels. At the highest gas flow rates sustainable by the experimental setup, the beginning of dry-out was observed intermittently in major portions of the channel, indicating transition from Annular to Dry flow.

4.2.5 Drry Flow

Higher gas flow rates forced the liquid film to the corners of the channels so that the liquid flowed only along the corners. As the gas flow rate was increased, the liquid flow rate decreased due to the increase of the hydraulic resistance while the driving pressure in the liquid chamber was steady. In the dry flow regime, the liquid flow rate decreased significantly.

Figure 4.6 Dry flow (D): gas superficial velocity (JG) ≈ 12.13 m/s, liquid superficial velocity (jL) ≈ 0.002 m/s and the liquid volumetric flow ratio (βL) ≈ 0.0016.

54 4.3 Gas-Liquid Two-Phase Flow Regime Maps

A flow map indicating the transitions between the major flow regimes is shown in Figure

4.7 including trend-lines for regime transitions identified from previous studies by Cubaud et al.

[19]. The current results are derived from channels made of PMMA, which is not molecularly smooth due to the specific micro-manufacturing processes used, and substantially less wetting than the glass and silicon used in the previous studies. Flow patterns were determined from images of steady state flow of each pattern. Through each set of experiments, the gas flow rate was gradually increased while the liquid flow rate determined, by the pressure in liquid chamber, was held constant. The transition between the Capillary Bubbly to Segmented-1 regimes occurred at approximately a fixed value of the liquid volumetric flow ratio βL ≈ 0.66. The

Segmented-1 to Segmented-2 transition occurred at βL ≈ 0.22, the Segmented-2 to Segmented-3

at βL ≈ 0.09, the Segmented-3 to the Transitional Segmented-Annular at βL ≈ 0.055, the

Figure 4.7 Gas-liquid two-phase flow map with regime separation lines for the flow in the serpentine test microchannels of the Type I AR=1 and Type II AR=2 chips.

Transitional Segmented-Annular to pure Annular at βL ≈ 0.018, and the Annular to Dry at βL ≈

0.003. Gas and liquid superficial velocities are plotted as a function of liquid volumetric flow ratio in Figure 4.8. As the gas superficial velocity increased, the liquid superficial velocity showed slight changes even when the applied pressure was steady. These liquid superficial velocity changes were caused by the changes in hydrodynamic resistance in the microchannel corresponding to each flow regime. The Segmented-1 and -2 regimes had increased hydrodynamic resistance. The higher gas superficial velocity also induced a lower liquid superficial velocity.

A flow map [103] for the directly micro-milled PMMA chips with aspect ratios of 1, 2 and 3 test channels [see Table 4.1 for dimensions] is shown in Figure 4.9 including trend-lines for regime transitions identified from the previous studies by Cubaud et al.[19], Günther et

56 Figure 4.9 Two-phase flow map for flow in serpentine test microchannels with different aspect ratios (AR) of mico-milled chips, (Dh=200 μm, +/-3%). Regime separation lines from previous works of Cubaud and Ho [19], Günther and Jensen [102], and Triplett et al. [104] (B-Bubbly, W-Wedge, S-Slug, A-Annular, WA- Wavy/Annular, AD-Annular/Dry, and D-Dry).

al.[21], and Triplett et al. [104]. A total of five microchips were used in producing the data in

Figure 4.9. Two with an AR of 1, two with an AR of 2, and one with an AR of 3; the chips will be denoted as AR1 and AR2 in subsequent discussion, with the subscript referring to the microchip number. This was done in order to examine reproducibility, given manufacturing and assembly induced uncertainties and to examine the possible effects of the presence of pressure taps on the results for the AR2 chips.

In the derived flow map the Segmented/Annular regime transition is along the same boundary for all of the aspect ratio channels used and in better agreement with the transition lines of Triplett et al. [104] and Günther et al. [21, 102]. The Bubbly to Segmented transition was

57 clearly demarcated for an AR of 1 and in relatively good agreement with that of Cubaud and Ho

[19]. No Bubbly regime was achieved for the AR of 3 channel; it can be assumed that the

Bubbly-Segmented transition for this aspect ratio does not quite coincide with the low aspect-

ratio channel. Similarly, in the channels with an AR of 2 a Bubbly/Segmented transition regime was observed only at the lowest combination of liquid and gas superficial velocities. The high

aspect ratio channels were narrower, since the channels had the same hydraulic diameter, which

would require injection of much smaller bubbles to create a Bubbly regime according to the definition. As a consequence, the bubbly regime transition line would be shifted to the left on the map. Certain flow map features may be as dependent on the injection characteristics as on a stable flow topology in the channel. The Bubbly-Segmented transition line from Günther et al.

[21, 102] is in disagreement with both the present data and Cubaud and Ho [19]; Günther et al.

[21, 102] obtained their data using a different liquid, ethanol, and channel materials, glass- covered PDMS, than Cubaud and Ho [19] and the present study. A different on-chip injection system was also used and discrete two-phase flows are known to be highly sensitive to initial conditions.

With respect to the three Segmented flow sub-regimes, it was observed that Segmented-2 is absent for the highest aspect-ratio channel. Transition from Segmented-1 to -2 was consistent for AR’s of 1 and 2, while the transition from Segmented-2 to -3 occurred at lower gas superficial velocities for an AR of 2 than for an AR of 1.

4.4 Details of the Segmented Flow Regimes

The segmented flow regime was further characterized as a function of gas bubble and liquid plug size and their regularity depending on conditions. In macro-scale studies of discrete gas-liquid flows, the bubble size distribution is strongly dependent on the initial and injection

58 conditions [105-107]. The same is true for the discrete regimes, Bubbly and Segmented flows, in

microchannels. The characteristics of the injection system strongly influence the characteristics

of the resulting discrete two-phase flow. A broad variety of discrete flow structures can be

achieved by controlled injection or pulsing, as demonstrated in a striking example on the mezzo

and macro scale by Knopf et al. [108]. In Segmented flow regime, the stability of the

configurations produced by the injection system under the influence of disturbances is critical.

Disturbances can originate from a variety of sources, including: (1) injection system feedback of the pressure fluctuations generated during pinch-off (release) of the bubbles; (2) feedback from downstream transitions, such as bubble release at the exit of the channel; (3) fluctuations in the feed rates induced from external perturbations; (4) the spatial variation of surface properties like wall roughness and wettability; (5) geometric changes in the channel such as bends, pressure taps, dimensional fluctuations, or variation of the cross-section; (6) decompression of gas bubbles along the channel; or (7) external vibrations. The third and last causes were ruled out by conducting the experiments on a vibration-isolated optical table (VH3660W-OPT, Newport

Corp., Irvine, CA) and with a stable and controlled fluid supply system. Considering the geometry, material and micro-manufacturing methods used, the fourth and fifth sources were present. The 1:4 - 1:16 cross-sectional area changes from the injection to the test channel were of particular significance, because of the ensuing expansion of the gas and the substantial inertia of the incoming liquid. The flow regime in the injection channel may be different from the one in the test channel, which would also cause local unsteadiness. Although the superficial velocities were much higher in the injection channel and maps for that small a hydraulic diameter are not available, this is not very likely because, the regime transition boundaries (Segmented-Annular,

Churn-Annular) are expected to be shifted to higher values of the superficial velocity as shown

59 by Zhao and Bi [44] and confirmed by Wälchli and Rohr [17]. For the other sources of disturbance, the first is a distinct possibility although the measured flow rates of both the liquid and the gas were quite steady, within ± 1% and ± 3% coefficient of variation (CV), respectively.

Feedback interaction with the gas path, whose exact volume and acoustic impedance were uncertain because of fittings, is a possibility, although long supply capillaries were used to increase the total impedance and reduce the potential contribution. The impedance of the on-chip liquid feed channels was at least an order of magnitude higher than that of the test section. The second disturbance source was also present, although the exit via was long and the sixth could still be a factor, although the pressure change in the test channel was kept modest, by limiting the flow rate window so that the end-to-end pressure change in the test channel was kept within ~3 psi, with a maximum estimated decompression volume change of less than 15%. The micro- milled AR of 1 Type I chip design was identical that of Cubaud and Ho [19], so it is reasonable to assume that it was subject to the same disturbance source characteristics.

Both regular and irregular Segmented flows were observed in both the micro-milled and hot embossed chips. The pattern was considered regular if it remained regular over the full length of the serpentine channel, which was more than 500 hydraulic diameters long. In keeping with

Günther et al. [102], regularity was based primarily on the length of the gas bubbles and not the

separating liquid plugs, which usually underwent much wider variations. A pattern was

considered regular if the coefficient of variation of the bubble length was less than 10%.

4.4.1 Image Processing

The measurements of gas bubbles and liquid plugs were made using image processing of

multiple images taken from the central area of the serpentine test section with a resolution of a

few micrometer per pixel listed in Table 4.2 depending on the objective and the binning size.

60 Table 4.2 Resolution of images acquired using objectives for different magnification and binning.

Magnification Binning Frame size µm/pixel Frame size

2X 1×1 1360 × 1036 pixels 3.301 ± 0.012 4.49 mm × 3.42 mm

2×2 680 × 518 pixels 6.625 ± 0.024

3×3 440 × 340 pixels 9.976 ± 0.053

4×4 332 × 256 pixels 13.274 ± 0.206

4X 1×1 1360 × 1036 pixels 1.633 ± 0.005 2.22 mm × 1.69 mm

2×2 680 × 518 pixels 3.268 ± 0.010

3×3 440 × 340 pixels 4.870 ± 0.017

4×4 332 × 256 pixels 6.438 ± 0.031

10X 1×1 1360 × 1036 pixels 0.654 ± 0.0001 0.89 mm × 0.68 mm

2×2 680 × 518 pixels 1.309 ± 0.004

3×3 440 × 340 pixels 1.957 ± 0.004

4×4 332 × 256 pixels 2.625 ± 0.006

40X 1×1 1360 × 1036 pixels 0.1634 ± 0.00 0.22 mm × 0.17 mm

2×2 680 × 518 pixels 0.3268 ± 0.0005

3×3 440 × 340 pixels 0.4895 ± 0.0007

4×4 332 × 256 pixels 0.6542 ± 0.0012

61 (a) (b)

(e) (f)

62 The images acquired using a CCD camera were 8-bit gray scale digital images, which

had 256 (0 for the complete white – 255 for the complete black) levels of gray from black to

white representing the intensity information of each pixel. Five steps in the image processing

using OPTIMASTM (ver. 6.51, Media Cybernetics, Inc., Bethesda, MD) software enabled

extraction of the geometry of the gas bubbles and liquid plugs. For the detection of pixels

representing a gas bubble, the edge lines of the microchannels needed to be removed. For this

process, a raw image (Figure 4.10 (a)) with gas bubbles and the channel edge line was divided by

a clearfield image (Figure 4.10 (b)) containing only the channel edge lines. This step produced the images (Figure 4.10 (c)) containing only gas bubbles. Figure 4.10 (d) illustrates the selected gas bubbles, outlined in red in Figure 4.10 (d), for the data extraction after the pixels inside the gas bubble were set to a value of 0 (complete white). Gas bubbles touching the image frame, outlined in green colour in Figure 4.10 (d), were excluded from the data extraction because only part of gas bubble was selected. For the detection of liquid plugs, another mask image (Figure

4.10 (e)) was added on the processed image ((Figure 4.10 (d)) to extract only the liquid plug between consecutive gas bubbles (Figure 4.10 (f)). Automating the image processing steps using scripts of commands enabled the automated processing of a number of images and provided higher statistical confidence by reducing human errors. Macro scripts used for extraction of gas bubbles and liquid plug lengths are described in Appendix B.

4.4.2 Gas Bubble and Liquid Plug Lengths

An overview of the variation of the measured average gas bubble and liquid plug lengths as functions of liquid volumetric flow ratio and Capillary number is given in Figure 4.11. The vertical error bars reflect the standard deviation of the length measured. The height of the vertical error bars relative to the average value is indicative of the extent of the irregularity of the

(a)

(b)

Figure 4.11 Scaled gas bubble and liquid plug lengths with respect to (a) the liquid volumetric flow ratio and (b) Capillary number for the Capillary bubbly and all Segmented flows in the microchannel of the hot embossed chips. Distribution of gas bubble length corresponding a’ to e’ are shown in Figure 4.12 (g).

64 associated Segmented flow. Regime boundaries identified from the flow map study of the hot embossed chip [see Figure 4.1] are indicated on the Figure 4.11 (a).

4.4.3 Reegularity of Segmented Flow

Examples of the typical regular and irregular Segmented flows from the microchannels of the hot embossed chips are shown in Figure 4.12 (a)-(e). The bubble size increased with lower

(a) (d)

(b) (e)

Figure 4.12 Representative distributions of the gas bubble length and images of air-water two- phase flows in the PMMA serpentine microchannels of hot embossed chip with Dh=200μm (nominal). Data points shown in (g), from (a’) to (e’), are corresponding to those in Figure 4.12 (a). (a) Capillary bubbly, βL=0.804, number of sample (bubble): 6,001, Range: 19.9µm, Mean: 152.61µm, Coefficient of variation (CV): 1.39%; (b) Segmented-1, βL=0.433, number of sample: 8,948, Range: 39µm, Mean: 235.24µm, CV: 2.76%; (c) Segmented-2, βL=0.26, number of sample: 6,897, Range: 64µm, Mean: 326.76µm, CV: 3.91%; (d) Segmented-2, βL=0.12, number of sample: 2,833, Range: 880.8µm, Mean: 816.88µm, CV: 15.5%; (e) Segmented-3, βL=0.068, number of sample: 1,246, Range: 2,838µm, Mean: 1,291.66µm, CV: 30.28%; (f) Observation region.

65 liquid volumetric flow ratio. In order to provide a more quantitative picture, representative distributions of the bubble length are shown in Figure 4.12 (g) corresponding to data points, (a’) to (e’) in Figure 4.11. Figure 4.12 (a), (b) and (c) showed regular flow patterns with coefficients of variation lower than 4%. The irregularity of the Segmented-2 and -3 flows of Figure 4.12 (d) and (e) is evident as a broader variation of the bubble lengths, with the coefficients of variation larger than 15%. The irregularity of the bubble length increased at higher gas flow rates, lower liquid volumetric flow ratios, even with stable gas and liquid injection flow rates.

One possible source of the increasingly irregular Segmented flow was the occasional coalescence of successive bubbles in the test channel section as the liquid plug became thinner due to the increased gas flow rate for a constant liquid flow rate. The other was the irregular generation of bubbles at the abrupt expansion of the channel due to gas expansion and the increased inertia of the liquid plug induced by the higher gas flow rate. These irregularities were discussed more specifically with Figure 4.15.

Regular Segmented flows at comparable bulk flow conditions are shown in Figure 4.13

(a) (b) (c)

Figure 4.13 Images of regular segmented air-water two-phase flow at comparable bulk flow conditions in PMMA serpentine microchannel (micro-milled chip with 3 aspect ratio test channels) with Dh=200 µm (nominal) for (a) AR=1, Segmented-2, βL=0.25, JL=46 mm/s; (b) AR=2, Segmenteed-2, βL=0.266, JL=43.4 mm/s; (c) AR=3, Segmented-3, βL=0.292, JL=46.2 mm/s.

66 for the three different aspect ratio channels at the lowest liquid superficial velocity examined.

There was an increase of gas bubble and liquid plug lengths when comparing flows from the lowest to the highest aspect ratio channels. This increase was large and can be only partially attributed to the ~20% larger cross-sectional area of the channel with an AR of 3 relative to that with an AR of 1. The increase in bubble length for the AR of 2 channel, which had a ~12% larger cross-sectional area, is less noticeable considering the variability of the bubble length, even for the regular Segmented flows. In order to provide a more quantitative picture, representative distributions of the bubble and liquid plug lengths are shown in Figure 4.14 for nearly the same bulk flow conditions and for all aspect ratios of the micro-milled chips. In all cases the Segmented flow was regular. There was a significant increase in both bubble and liquid plug lengths as the aspect ratio increased to an AR of 3. The differences were less remarkable between the AR of 1 and AR of 2 channels. The distributions of the liquid plug lengths were

(a) (b)

67 broader than those of the bubble length, confirming a greater variability in the liquid plug lengths, even for regular Segmented flow.

The irregularity of the segmented flows in the Type II chips was substantially less than

that observed in the same aspect ratio Type I chips. This was most likely because of the

impedance of the air supply channel on the Type II chips was six times larger than that of Type I,

while the channel wall roughnesses were comparable. This confirms that feedback from supply

(a) (b)

68 lines, particularly that of the gas, was one of the principal causes of Segmented flow irregularity.

However, increasing the impedance of the on-chip gas supply line by almost an order of magnitude, with the associated substantial increase in chip footprint, did not eliminate irregularity [see Figure 4.11]. In addition to the supply impedance, geometric differences and spatial variations in surface properties also made significant contributions.

Two distinct scenarios in the development of irregularity were observed. These are illustrated in Figure 4.15 based on evidence from micro-milled chips. Under Bubbly/Segmented flow conditions, the injection produced fairly regular gas bubble lengths. However, these tended to develop uneven length separating liquid plugs. This resulted in coalescence of the injected bubbles into longer bubbles as they approached each other [see Figure 4.15 (a)]. This coalescence involved two or more bubbles as shown by the measured multi-modal bubble length distribution shown in Figure 4.15 (b). The higher modes of this distribution are not exact integer multiples of the lowest mode, which is consistent with the doubling of the volume not resulting in exact doubling of the length, because of the bubble end-cap geometry. The same phenomenon was also observed at lower liquid volumetric flow ratios, at which the Segmented flow became irregular through coalescence, as shown in Figure 4.16 (b). This occurred after the original

Segmented flow was regular over a much longer channel length. This coalescence-induced irregularity was often intermittent as seen in Figure 4.16 (a), where the regular flow was sustained throughout the channel during the same experiment as the flow shown in Figure 4.16

(b).

Small perturbations in the injected bubble volume and/or in the volume of the separating liquid plug, lead to mutual pair-wise approach of successive bubbles and coalescence. Taking into consideration the Taylor-type recirculation flow patterns documented in Segmented flows

(a) (b)

Figure 4.16 (a) Regular and (b) irregular segmented flow through coalescence at the same bulk flow conditions AR=1, βL=0.415, JL=57.7 mm/s.

through μPIV by Günther et al.[22], development of liquid plugs that are strongly uneven

between the leading and trailing ends of a bubble can result in a net drafting effect on a trailing

bubble and concurrent slowing of a leading bubble. This can bring the bubbles closer together

until coalescence becomes possible through the elimination of the dividing liquid plug/film.

Bends can also play a role in inducing the perturbations needed to trigger the phenomenon, partially because of the resulting break in symmetry of the liquid plug flow [see Günther et

al.[21]], and because of inhomogeneities introduced in the liquid films/corner flows around the

bubbles as they pass through the bends. This mechanism of irregular Segmented flow

development through coalescence was observed predominantly in Segmented-1 and Segmented-

2 topologies, and can also be affected by wall surface properties, which influence contact angles,

when they exist, and the viability of liquid films between the bubbles and the channel walls.

The second scenario for the development of irregular Segmented flow portrayed in Figure

4.15 (c) is due predominantly to unsteady injection, where usually long bubbles of uneven size

are continuously generated at the exit of the injection channel leading to single-mode bubble

length distributions as that of Figure 4.15 (d) with no corresponding evidence of coalescence.

70 This typically occurred for Segmented-3 flows at lower liquid volumetric flow ratios, the

majority of which were irregular. The influence of unsteady injection on irregular Segmented

flow generation is also seen in Figure 4.12 (e). Apart from feedback possibly affecting the gas-

liquid flow generation at the upstream cross junction, the entry into the test channel could be

irregular because of abrupt gas expansion and high liquid plug inertia coming from the much

faster flow in the mixing channel, which enters the test channel through the abrupt, asymmetric

expansion.

The variability of the liquid plug length was substantially higher than that of the bubble

length, even in the cases with regular flow [see Figure 4.11]. This is consistent with Günther et al.

[102], who showed > 30% variation in liquid plug lengths. The variation remained despite their introduction of large on-chip feed-line impedances, to help stabilize the incoming gas flow rates and regularize the size of the bubbles at the cost of increasing the total chip pressure drop by an order of magnitude. The physical explanation for this increased variability is that, unlike the gas in the bubble, liquid can drain from preceding to following (and vice versa) plugs, primarily through corner flows, which connect them. Changes of bubble size and shape, because of decompression or channel geometry changes such as the bends, can induce transfer between successive liquid plugs.

The liquid plug length variation was not monotonic at all liquid superficial velocities. It decreased as the gas content of the flow increased, reaching a minimum near a βL of 0.2 and then increased again. No systematic difference in the average bubble and liquid plug lengths was apparent between an AR of 1 and an AR of 2 outside the variability margins of these quantities, for all liquid superficial velocities. This was independent of whether the Segmented flow was regular or irregular. However, a significant effect was observed after increasing the aspect ratio

71 to an AR of 3 [see Figure 4.13] at the lower liquid superficial velocities. Both the gas bubble and

liquid plug lengths were much higher at an AR of 3 than those measured at the lower aspect ratios for almost the full range of the liquid volumetric flow ratio over which the Segmented flow was observed.

4.5 Pressure Drop in Microchannels

4.5.1 Single Phase Frictional Pressure Drop

The analytical solution of the classic Poisson equation (Equation 4.1) for the volumetric

flow rate of the fully developed laminar flow in the rectangular channel was reviewed by Berker

(1963) [109] and shown in Equation 4.2.

1 ̂ (4.1)

4 ∆ 192 tanh/2 1 (4.2) 3 ∆ ,,, where Q is the volumetric flow rate, µ is the dynamic viscosity of the fluid, a and b are the half

length of the channel width and depth, respectively.

In the linear case, the fluid resistance (R) is related to the pressure drop and volumetric

flow rate as [110];

∆⁄ (4.3)

By substituting Equation 4.2 into 4.3, the fluid resistance (R) in a rectangular channel

becomes Equation 4.4.

3∆ 192 tanh/2 (4.4) 4 1 ∑ ,,,

72 and the pressure drop per unit length of channel can be written as Equation 4.5:

∆ 3 ∆ 192 tanh/2 (4.5) 4 1 ∑ ,,,

Another analytical solution (Equation 4.6) of Equation 4.1 for the pressure drop per unit length of laminar flow in a rectangular channel was derived by Shah and London (1978) [111].

∆ 1⁄ 241 1.3553 1.9467 1.7012 0.9564 0.2537 ∆ 8

(4.6) where α is the aspect ratio defined as the channel depth divided by the channel width, V is the

characteristic velocity of the fluid, and µ is the dynamic viscosity of the fluid. Variation of

values from Equations 4.5 and 4.6 is less than 0.5%.

The Darcy-Weisbach equation (Equation 4.7) [112] is widely used for the calculation of

pressure drop due to friction in a channel.

∆ (4.7) ∆ 2 where f is the Darcy friction factor, Δ is the channel length, is the hydraulic diameter, is

the density of fluid and is the characteristic velocity of fluid in microchannel. The value of the

friction factor for laminar flow in a channel is obtained from Equation 4.8.

64 64 (4.8)

∆ 32 (4.9) ∆

73 In case of fully developed laminar flow in a square microchannel with an aspect ratio, α = 1,

Equations 4.5, 4.6 and 4.7 provided the same final equation (the Hagen-Poiseuille equation,

Equation 4.9).

4.5.2 Homogeneous Flow Model for the Two-Phase Flow Pressure Drop

One of the models for the two-phase flow pressure drop is the homogeneous flow model, which was developed from the Darcy-Weisbach equation (Equation 4.7) by using a homogeneous mixture density () and an effective dynamic viscosity (). The modified

Darcy-Weisbach equation is defined as shown in Equation 4.10:

Δ 1 ΔZ (4.10) 2

where G is the total mass flux, , and is the homogeneous mixture density defined as Equation 4.11:

1 1 (4.11)

where is the density of the gas and is the density of the liquid, and x is the mass quality, which is a ratio of gas mass flow rate to total mass flow rate. The is the two-phase Darcy

friction factor (Equation 4.12):

a⁄ (4.12)

where a is the fitting parameter determined from experiment and is the two-phase Reynolds

number,

74 (4.13)

where G is the total mass flux, , and is the effective dynamic viscosity.

Previous two-phase effective dynamic viscosity models are,

McAdams (1954) [113]:

1 (4.14)

Cicchhitti et al. (1960) [114]:

1 (4.15)

Owens (1961) [115]:

(4.16)

Dukler et al. (1964) [116]:

1 (4.17)

Beattie and Whalley (1982) [117]:

11 2.5 (4.18)

Lin et al (1991) [118]:

. (4.19) where is the mass quality, i.e. the ratio of the gas to total mass flow rate.

75 4.5.3 Separated Flow Model for the Two-Phase Flow Pressure Drop

The correlation of Lockhart and Martinelli [48], also called the separated flow model,

was developed to estimate two-phase flow pressure drops. The two-phase flow frictional

multiplier and the Lockhart-Martinelli parameter (X) were defined as shown in Equations

4.20 and 4.21, respectively.

∆ ∆ (4.20) ∆ ∆

∆ ∆ (4.21) ∆ 1 ∆ where Δ⁄ Δ is the frictional pressure drop per unit length, subscripts TP, L and G denote

two-phase, liquid and gas flow, is a viscosity ratio of liquid and gas ⁄ and is a liquid

volumetric flow ratio. The relation of the friction multiplier and parameter (X) is the inverse second-order polynomial as shown in Equation 4.22 [119].

1 1 (4.22)

Table 4.3 Parameter C in Lockhart-Martinelli correlation (Chisholm, 1967).

Fluid phase C Liquid Gas

Turbulent Turbulent 20

Laminar Turbulent 12

Turbulent Laminar 10

Laminar Laminar 5

76 where C is a parameter depending on whether the liquid and gas phase are turbulent or laminar, as shown in Table 4.3. A C value of 5 was used for the Equation 4.22 in case both liquid and gas flows in the microchannels were laminar.

While Chisholm proposed the C values for larger channels, Mishima and Hibiki [120] have proposed a correlation of the parameter C for smaller channels in terms of the hydraulic diameter as shown in Equation 4.23.

2110.319 (4.23)

where the Dh is hydraulic diameter of microchannel.

Kawahara et al. [24] obtained the value of parameter C of 0.24 by averaging experimental data and showed good agreement with the result of Lee and Lee’s model [49].

Figure 4.17 Two-phase frictional multiplier in terms of Lockhart-Martinelli parameter (X).

77 Figure 4.17 shows each C values and the two-phase flow multiplier as a function of the

Lockhart-Martinelli parameter from Chisholm [119], Mishma and Hibiki [120] and Kawahara et

al [24].

4.5.4 Measurements of Gas-Liquid Two-Phase Flow Pressure Drop

The gas-liquid two-phase flow pressure drops in microchannels of the hot embossed

PMMA chip were measured and scaled by the corresponding liquid single phase flow pressure

drop. The single phase flow pressured drops were calculated using the Hagen-Poiseuille equation

[Equation 4.9] derived from the Navier-Stokes equations for fully developed laminar flow in channels. The equation relates the pressure drop (∆) along the channel length (∆) for velocity

(), dynamic viscosity () and hydraulic diameter ().

Figure 4.18 Two-phase frictional multiplier ( ) in terms of Lockhart-Martinelli parameter () with constant C=1.39 for AR=1 test channels. CB: Capillary Bubbly, S1: Segmented-1, S2: Segmented-2, S3: Segmented-3, SA: Segmented-Annular, A: Annular and D: Dry flows.

78 The correlation of Lockhart and Martinelli [48], also called the separated flow model, was developed to estimate the gas-liquid two-phase pressure drops by applying a frictional

multiplier ( ) to the liquid single phase flow pressure drop. The multiplier is correlated in terms of Lockhart-Martinelli parameter (), which are defined as Equations 4.20 and 4.21. The relation between the multiplier and the parameter is the inverse second-order polynomial as shown in Equation 4.22. Values of C were estimated by several groups [24, 119, 120] based on different experimental conditions; these inverse second-order polynomials are plotted in Figure

4.18 along with the measured pressure drops.

A value of C of 1.39 gave relatively good agreements between the experimental data and the correlation in the Bubbly, Segment/Annular, Annular and Dry flow regimes, which were also

Table 4.4 Variables for the abscissa and ordinate used in Lockhart-Martinelli correlation and this work.

Lockhart-Martinelli (1949) This work

∆ ∆ ∆ 1 ∆ Abscissa

∆ ∆ ∆ ∆ ∆ Ordinate ∆ ∆ ∆

79 observed in macro-scale channels. However, in the Segmented flow regimes, which in microchannels are dominated by interfacial forces over inertial forces, the experimental data pressure drops were significantly higher than predicted by the correlation. Transforming the independent variable from the Lockhart-Martinelli parameter () to the liquid volumetric flow ratio () [see Table 4.4] and Capillary number (Ca) clarifies the pressure drop trends in each flow regime [see Figure 4.19 and Figure 4.20]. In the Capillary/Bubbly flow regime, the scaling factor was very close to unity, showing that the gas bubbles were not in contact with the channel walls and did not contribute significantly to the pressure drop. The initial increase of the two- phase pressure drop relative to the correlation was in the Segmented-1 sub-regime, with a local maximum under Segmented-2 flow, and a drop off and local minimum corresponding with the

Segmented-3 regime. With the transition to Segmented-Annular, Annular, and Dry flows the pressure drop resumed increasing and following the trend estimated by the Lockhart-Martinelli relationship. Variation of the two-phase pressure drops in the three Segmented flow sub-regimes directly correlates with the distinctly different flow behaviours. The criteria used to distinguish between these regimes are partially due to changes in thin film or corner liquid flows.

Understanding the changes occurring in the local liquid flow characteristics due to the increased hydraulic resistance with higher gas flow rates as the volumetric flow ratio changes [see Figure

4.8] is essential for explaining the behaviour. Transitions between these sub-regimes are related to changes between the flow states where bubbles are pushed through the channel much like blockages, flow states where a fraction of the liquid bypasses the bubbles through liquid film and corner flows between bubbles and walls, and flow states where the liquid films have disappeared through instability and only corner flows remain. In a square cross-section microchannel, liquid flows through the liquid film and corner flows between the bubble and channel wall.

(a)

(b)

Figure 4.19 Scaled two-phase pressure drop as a function of liquid volumetric flow ratio (βL) for (a) all flow regimes and (b) details of Segmented flow regime.

(a)

(b)

Figure 4.20 Scaled two-phase pressure drop as a function of Capillary number (Ca) for (a) all flow regimes and (b) details of Segmented flow regime.

82 The principal mechanism for dissipating the pressure-driven force was the liquid films

lubricating the bubbles. The largest pressure drop occurred at the leading edge of each bubble

[121]. Kreutzer et al. [39, 122] investigating pressure drops across gas bubbles in capillary tube identified static head loss due to the presence of bubbles, which is proportional to the number of

bubbles per unit length and inversely proportional to the liquid plug length, as another

contribution to pressure drop in addition to viscous wall friction losses. The more bubbles per

unit channel length, the higher the pressure drop in the microchannels. This observation is

consistent with the experimental data as well. Figure 4.21 shows the number of bubbles,

alternatively the bubble frequency, present in the channel between the two pressure ports. The

number of bubbles increased linearly in the Bubbly

83 flow regime and attained the highest frequency during the Segmented-1 sub-regime before

decreasing through the Segmented-1 and Segmented-2 sub-regimes. The liquid plug length decreased in the Segmented-1 sub-regime, remained steady in the Segmented-2 sub-regime, and increased slightly in the Segmented-3 sub-regime [see Figure 4.11].

In addition to this relationship between the two-phase flow pressure drop and the number of bubbles, the proportional length of the gas bubble and the liquid plug (i.e., proportion of viscosities of gas and liquid) also play a role in pressure drop per unit length. In a homogeneous flow model [104] for the two-phase flow frictional pressure drop [see section 4.5.2], the dynamic viscosity of the fluid in Equation 4.9 is replaced by the effective viscosity of the gas and liquid.

While several groups [see Equation 4.13 – 4.18 [113-118]] have defined their own values

based on experimental data, it can be simply represented as a function of the dynamic viscosities

of the gas and liquid and the mass quality (x), which is the ratio of the gas flow rate to the total

mass flow rate. The homogeneous flow model is the simplest estimate of the frictional two-phase

pressure drop contributed by viscous frictional energy dissipation. Although this model does not

agree well with flow in microchannels, it illustrates the dependence of the pressure drop on the

fraction of gas and liquid per unit length. Provided the same flow velocity and cross-sectional

area of channel, the contribution of the liquid flow to the pressure drop is much higher than that

of the gas flow. This may also explain the reduced pressure drop even at higher flow rates in the

Segmented-2 and Segmented-3 sub-regimes showing the relatively longer gas bubble. Even

though the gas flow occupied most of the channel in Segmented/Annular, Annular and Dry flow

regimes, two-phase flow pressure drops increased due to the much higher gas superficial

velocities in these flow regimes and the reduced cross-sectional area due to the confinement of

liquid to the channel corners.

84 4.6 Conclusions

Fundamental aspects of gas-liquid two-phase flows in microchannels fabricated on poly-

methyl methacrylate (PMMA) were investigated experimentally. The PMMA channel walls were

partially non-wetting, with a typical static water contact angle of 65o in stock form, and not

molecularly smooth with roughnesses on the order of a few hundred nanometers. Two-phase

flow regimes were subject to a combination of gas and liquid volumetric flow rates. Flow maps were developed base on observed flow regimes at the specific gas and liquid superficial velocities. Variations of the measured average gas bubble and liquid plug lengths as a function of liquid volumetric flow ratio were overviewed. Two main mechanisms were identified as responsible for the development of irregular Segmented flow under steady gas and liquid flow rates. One is related to the progressive mutual approach of bubble pairs along the channel, resulting in coalescence, and induced by small perturbations. This mechanism occurs mostly at relatively high liquid volumetric flow ratios. The other mechanism is attributed to unsteady injection because of abrupt gas expansion and high liquid plug inertia at the transition from a smaller injection channel to the test channel with relatively low liquid volumetric flow ratios.

The variability in the size of liquid plugs separating gas bubbles in Segmented flow is found to be substantially higher than that of the bubbles even when the flow is regular with a low variability of bubble size. Three distinct Segmented flow sub-regimes were identified with different gas bubble lengths, liquid plug lengths and the number of gas bubbles. Distinct variations of the two-phase pressure drops were associated with each of two-phase flow regimes.

Pressure drops in gas-liquid two-phase flow in microchannels are functions of the superficial velocities and lengths of the gas bubbles and liquid plugs, the flow regimes, the number of bubbles across the channel.

85 Chapter 5 Liquid-Liquid Segmented Flows in Microchannels

5.1 Introduction

Liquid-liquid segmented flows in microchannels have drawn attention because of their

potential advantages over single phase flows in biochemical analytical devices offering improved

heat and mass transfer [123] with higher surface to volume ratio. Encapsulated dispersed fluid in

the shape of spherical droplets or elongated plugs also serve to confine the dispersion of target

molecules [21, 51], which are normally widely dispersed in pressure-driven single phase flow.

Interaction of shear and interfacial forces on the interface between the dispersed and carrier

fluids induce recirculation of streamlines inside both the dispersed fluid and the carrier fluid

plugs, intensifying mixing [51] which is reduced by the laminar characteristics of microchannel

flows with low Reynolds numbers (Re << 10). The presence of continuous thin films of carrier

fluid between the dispersed one and the channel walls can prevent target molecules suspended in

the droplets from being adsorbed on the channel surface. This is the only potential source of

cross-contamination between successive dispersed plugs containing different reagent mixtures.

Segmented flows have been demonstrated in many practical applications including,

continuous segmented flow polymerase chain reaction (PCR) [9], encapsulation of enzymes in

lipid vesicles [124] using microfluidic jetting, protein crystallization [52], and fabrication of

magnetic hydrogel microparticles [125]. The manipulation of the dispersed droplets and plugs

[12, 51, 126] were studied in order to replace conventional micro titer plates with a series of

continuous separated droplets in microchannels for high throughput screening in drug discovery

[10]. Each encapsulated droplet served in the designed function as an independent microreactor

with increased mass and heat transfer and maintained the distinct identity of the contents without

cross communication between neighboring dispersed fluid volumes.

86 5.1.1 Configurations of Microfluidic Test Chips

The schematic configurations and characteristic dimensions of the microchannels fabricated on the polycarbonate chips are shown in Figure 5.1 and Table 5.1. Three different expansion ratios

from the injection to the test channel were used to observe the effect of different cross-sectional

areas of the injection channel on the generation of segmented flows in the test channel. Carrier

and dispersed fluids were introduced into each inlet and meet at a cross junction channel. Carrier

fluid from the sides periodically pinched the elongated cylindrical dispersed fluid thread from the central branch. Shear and interfacial forces at the interface between the dispersed and carrier fluids generated continuous mono-dispersed droplets in the central branch past the cross junction

[50, 57]. This two-phase flow entered the test channel, which was expanded deeper (Type I) and wider (Type I, II and III) in a serpentine configuration. Two 50 µm square channels for pressure

taps, which enabled measurement of the pressure drop across the test section, were located on

either end of the serpentine test channel. The serpentine configuration was chosen due to its

compact footprint and also because it allows simultaneous observation of flow over almost the

full channel length. The total channel length between two pressure tap channels was ~120 mm.

The U- and 90o- bends in the path of the test channel had a nominal centreline radius of 105 µm

(±6%) for all chips. Three different expansion ratios from the injection to test channels were

evaluated.

The cross-section dimensions of the injection channels were 50 μm × 50 μm for Type I,

50 μm × 200 μm for Type II, and 100 μm × 200 μm for Type III with a fixed test channel cross-

section of 200 μm deep × 200 μm wide. The cross-junction of the injection section and

expansion area from the injection to test channels had rounded corners resulting from the finite

cutting tool radius (r) of 100 μm, as shown in Figure 5.1 (b).

(a)

(b)

(c)

88 Table 5.1 Characteristic dimensions of the injection and test channels

Test chip Injection channel Test channel Expansion ratio

Type I 50 (w) µm × 50 (d) µm 1:16

Type II 50 µm × 200 µm 200 µm × 200 µm 1:4

Type III 100 µm × 200 µm 1:2

5.2 Properties of Test Fluids

Deionized water (18.2 MΩ·cm at 25°C, TOC of 5 ppb, Millipore, Billerica, MA) with 1 %

(v/v) blue food-coloring (McCormick, Sparks, MD) was used as the dispersed phase fluid.

Perfluorocarbon (perfluorotripropylamine, FC 3283, 3M, St. Paul, MN) with 10% (v/v) nonionic fluoro-soluble surfactant, 1H, 1H, 2H, 2H-Perfluorooctanol, 97% (PFO, Alfa Aesar, Ward Hill,

MA), was used as the carrier phase fluid. Measurements of the physical properties of the test fluids like wettability, density, interfacial force and viscosity were essential for the analysis of the experimental results. Density of the carrier and dispersed fluids was measured using an electronic scale (B120S, Sartorius Corp, Edgewood, NY). Contact angle of test fluids on the

PMMA and PC substrates and interfacial forces between two immiscible fluids were measured from a pendant water drop suspended in carrier fluid [127] using a contact angle and surface tension measurement instrument (FTA125, First Ten Angstroms, Inc., Portsmouth, VA), and viscosity was measured using a calibrated U-shaped glass tube viscometer (9721-B50, Cannon

Instrument Co., State College, PA).

89 5.2.1 Wettability and Surfactant

Perfluorocarbon was chosen for the carrier fluid because its high wettablility and

compatibility with polymers, which are the principal required characteristics for the carrier fluid.

In liquid-liquid segmented flow in a microchannel, one of fluids wetted the channel wall dominantly and encapsulated the other fluid as the segmented flow. Higher wettability of the carrier fluid than that of the dispersed fluid was required to maintain stable segmented flow.

Comparing other fluids used for the carrier fluids like glycerin, mineral oil, dodecane and hexadecane, relatively lower viscosity was another reason to choose the fluorocarbon as a carrier fluid. Contact angles of FC 3283 + PFO (10% v/v) on the polymer (PMMA and PC) substrates are illustrated in Figure 5.2 (a) and (b) as an inverse measure of wettability. Comparable

examples showing lower wettability and higher contact angles for deionized water on those

substrates are also given in Figure 5.2 (c) and (d). While the contact angles of the deionized

water on the substrates ranged between 65° and 85°, the fluorocarbon fluid spread out on the

polymer substrates and showed complete wetting with undetectable contact angles.

A surfactant, 1H, 1H, 2H, 2H-Perfluorooctanol (PFO, CF3(CF2)5(CH2)2OH), was used to

reduce the surface tension of the carrier fluid and the interfacial force between the dispersed and

carrier fluids. PFO has a hydrophobic (i.e., fluorophilic in this study) tail and a hydrophilic head

as illustrated in Figure 5.3 (b). Therefore, PFO is soluble in both carrier and dispersed fluids.

Water droplets sat on the substrate under equilibrium static conditions validate the Young’s relation ( ) between interfacial forces working on the three interfaces,

water/substrate, fluorocarbon/substrate and water/fluorocarbon, [see Figure 5.3 (a)]. Surfactant

resolved in both fluids reduced the surface tension of the carrier fluid (γSO) and the interfacial

force between the carrier and dispersed fluid (γWO), [see Figure 5.3 (c)]. In order to improve the

90 PMMA PC

(a) (b)

PMMA PC

wetting of the fluorocarbon carrier fluid on the channel walls and prevent the dispersed fluid from wetting the microchannel wall, the microchannel walls were modified using the following process: Vaporized 1H, 1H, 2H, 2H-perfluorodecyltrichlorosilane (ABCR GmbH & Co. KG,

Germany) was driven by compressed argon (Ar) gas into the microchannels for 1 hour and then the microchannel was filled with FC 3283 + PFO (10% v/v) solution and baked in a convection oven at 60 °C for 1hr. The surface treatment decreased the wettability of the water on the substrate by increasing the surface tension between the water and substrate, γSW and increased the wettability of the fluorocarbon fluid on the substrate decreasing the surface tension between the fluorocarbon fluid and substrate, γSO, [see Figure 5.3 (d)].

(a) (b)

Under dynamic conditions driven by a pressure gradient, Figure 5.4 shows different dynamic contact angles between the carrier fluid and channel wall in liquid-liquid segmented flows (a) with and (b) without surfactant in the carrier fluid. With the increased wettability of the carrier fluid and decreased contact angle between the carrier fluid and channel walls, there exits carrier fluid thin film between dispersed fluid and walls as a function of characteristic velocity of dispersed fluid. Bretherton [38] developed a lubricating theory explaining that the thin film between dispersed fluid and channel wall increases with the characteristic velocity of the

92 (a) (b)

dispersed fluid. The thin film works as a barrier preventing the wetting of the dispersed fluid on

the channel walls.

5.2.2 Measurement of Viscosity

Viscosity of a fluid is the resistance to flow due to internal friction between layers of fluids

and one of the main parameters contributing to the increasing frictional pressure drops in

microchannels. Therefore, selection of a fluid with a lower viscosity is always preferred to keep

the working pressure low for practical applications. A calibrated glass capillary viscometer using

ASTM D 445 and ISO 3104 (Cannon-Fenske Routine Viscometer, Cannon Instrument Co., State

College, PA) was used to measure the viscosities of the test fluids. After measuring the efflux

time for the meniscus to pass from mark E to mark F on the viscometer as shown in Figure 5.5,

the kinematic viscosity in mm2/s (cSt) was calculated by multiplying the efflux time in seconds

by the viscometer constant and the dynamic viscosity in mPa·s (cP) was acquired by multiplying

the kinematic viscosity by the density. The viscometer constant was given as a 0.0022028 mm2/s2 (cSt/s) at 22.5 °C.

Figure 5.5 Callibrated Cannon-Fenske Routine Viscometers according to ASTM D 445 and ISO 3104.

5.2.3 Measurement of the Surface Tension

The pendant drop method was used to measure the interfacial force between the carrier

and dispersed fluids using a contact angle and interfacial tension measurement instrument

(FTA125, First Ten Angstroms, Inc. Portsmouth, VA), [see Figure 5.6 (a)]. Pendant drops

distorted by gravity are balanced by the interfacial force between the carrier and dispersed fluids.

The interfacial forces are determined by fitting the profile of the drop to the Young-Laplace

equation, [see Equation 5.1].

∆ / (5.1) where ∆ is the mass density difference between the droplet and the surrounding fluid, is the gravitational constant, is the radius of curvature at the apex of the drop [see Figure 5.6 (b)],

94 and is the shape factor of droplet and calculated from the ratio ⁄ . Hansen and

Rødsrud [127] calculated β by using polynomial regression analysis of σ-data produced theoretical profiles as shown in Equation 5.2.

0.12836 0.7577 1.7713 0.5426 (5.2)

The contact angle and interfacial force measurement instrument extract σ data automatically by using image processing on the captured images.

Figure 5.7 shows representative images of water pendant drops suspended in FC 3283 with different volumetric concentrations of surfactant.

(a) (b)

Figure 5.6 (a) Surface tension and interfacial measurement system (FTA125) (b) Water pendant drop suspended in air with variables representing drop shape.

(a) (b)

All of the fluidic properties including density, viscosity, wettability, and interfacial force of the test fluids were measured more than 10 times at room temperature (22°C ± 1) with less 3% coefficients of variation of them, [see Table 5.2 and Figure 5.8]. With the volumetric concentrations of surfactant, density of the carrier fluid and the interfacial force between the carrier and dispersed fluids decreased, but viscosity of the carrier fluid increased.

5.3 Liquid-Liquid Segmented Flow Regimes

Three distinct liquid-liquid segmented flow regimes, Droplet, Plug, and transient

96 Table 5.2 Properties of the dispersed and carrier fluids. DI-water and perfluorocarbo (FC3283) with 10% (v/v) nonionic surfactant (PFO, Perfluorooctanol) were used as test fluids.

3 ρ (Kg/m ) γOW (mN/m) µ (mPa·s, cP)

Di-water 1) 997.83 ± 0.72 - 2) 0.9509 ± 0.0059

FC 3283 1,808.58 ± 9.46 54.15 ± 0.13 1.4796 ± 0.0053

FC3283 + PFO (5% v/v) 1,789.87 ± 8.65 14.79 ± 0.03 1.6007 ± 0.0079

FC3283 + PFO (10% v/v) 1,777.54 ± 8.06 13.49 ± 0.33 1.7816 ± 0.0121

FC3283 + PFO (20% v/v) 1,767.03 ± 10.54 12.25 ± 0.42 2.2475 ± 0.0115

* All properties were measured at room temperature (22.5 ± 1 °C) 1) 997.46 Kg/m3 at 22.5°C, F. M. White, Fluid Mechanics, 5th edition, 2003, McGraw-Hill. 2) 0.9534 mPa·s at 22.5°C, F. M. White, Fluid Mechanics, 5th edition, 2003, McGraw-Hill.

(a) (b) (c)

97 Irregular Segmented flows were observed in the test channels. Specific flow regimes at the

expansion area from the injection to the test channels and at the test channel sections were

specified based on their topology in terms of the carrier fluid volumetric flow ratio, which is the

fraction of the volumetric flow rate of the carrier fluid in the total volumetric flow rate of the

carrier and dispersed fluids, ⁄ , where and are the volumetric flow rates

of the carrier and dispersed fluids, respectively.

5.3.1 Type I Chip with an Expansion Ratio of 16

The images in the first column in Figure 5.9 (a), (c), (e) and (g) were acquired from the

expansion area of the Type I chip with an expansion ratio of 16 using a 4X objective under white field illumination for quantitative extraction of geometrical data. The second column of Figure

5.9 (b), (d), (f) and (h) shows images acquired from the test channel through a 10X objective

with double-pulsed 532 nm laser illumination for the measurement of the velocity of the

dispersed droplets and plugs. Figure 5.9 (a) and (b) show regular water droplets flowing along

the centerline of the channels. With increasing flow rate of the dispersed fluid the distance

between the neighboring droplets was reduced, increasing the frequency of droplet injection

(Figure 5.9 (c)). Some of droplets in the test channels coalesced producing an Irregular

Segmented flow regime (Figure 5.9 (d)) due to disturbances like the variation of the surface

roughness and the curvature effects resulting from the serpentine test channels. During the

repeated 90° or 180° turns in the serpentine test channels, the velocity profile changed and the

centrifugal forces acting on the droplets varied accordingly. Scattered Droplet flow [see Figure

5.9 (e)] was observed in the expansion area with increased dispersed flow rate only in the Type I

chip with the highest expansion ratio. Intimate contact between the scattered droplets resulted in

coalesced larger irregular droplets in the test channels with the same reason described above.

(a) (b)

(e) (f)

(g) (h)

(a) (b)

(e) (f)

(g) (h)

Figure 5.10 Liquid-liquid segmented flow regimes of the 50 µm width × 200 µm depth injection channel chip (Type II chip with an expansion ratio of 4) under white field illumination (a) Droplet flow at the injection channel (4X objective) with homogeneous carrier fluid volumetric flow ratio, βC ≈ 0.87 (b) Droplet flow in the test channel (2X objective) with βC ≈ 0.87 (c) Irregular Segmented flow, βC ≈ 0.69 (d) Irregular Segmented flow, βC ≈ 0.67 (e) Plug flow, βC ≈ 0.4 (f) Plug flow, βC ≈ 0.4 (g) Plug flow, βC ≈ 0.11 (h) Plug flow, βC ≈ 0.11.

100 Unlike instantaneous coalescence of the contacting gas bubbles observed in gas-liquid two-phase flow in microchannels [128], dispersed aqueous droplets in the perfluorocarbon carrier fluid maintained their own interfaces for a while even as they adhered closely to each other.

Below 0.32, Figure 5.9 (g) and (h) show regular Plug flows with sufficient distance between the plugs to prevent the dispersed liquid plugs from coalescing.

5.3.2 Type II Chip with an Expansion Ratio of 4

Images in the first and second columns in Figure 5.10display segmented flow at the expansion area and test section of the Type II chip with an expansion ratio of 4 through 4X and

10X objectives under white illumination, respectively. Droplet, Plug, and transient Irregular

Segmented flows were also observed. Unlike irregular flow regimes developed by coalescence of neighboring droplets (Figure 5.9 (c)) and scattered droplets (Figure 5.9 (e)) on the Type I chip, irregular flow regimes on the Type II chip were contributed by both coalescence of neighboring droplets and irregular injection in the expansion area.

5.3.3 Type III Chip with an Expansion Ratio of 2

Figure 5.11 shows segmented flow regimes from the Type III chip. Most of the segmented flow regimes observed on this chip were regular Plug flows without a transient

Irregular Segmented flow regime, for the carrier and dispersed fluids flow rates used in this observation window. Due to the relatively larger injection channels with the lower expansion ratio, more stable and longer plugs were generated in the test channel. Figure 5.11 also shows the dependence of the length of dispersed plug, the distance between dispersed plugs, and the residence time (i.e., frequency) of the dispersed plugs on the different combinations of carrier and dispersed flow rates.

It is evident that for the cross junction injector the injection channel’s cross-sectional area

101

(a) (b)

has a strong influence on the two phase flow regime generated in the test channel. Smaller

injection channel cross-sectional areas favor the formation of droplet flow regimes, and when

these regimes are created at higher volumetric flow ratios they are prone to an unstable outcome

because of closer packing of the larger numbers of generated droplets. Larger injector cross-

sectional area channels tend to produce primarily plugs. These trends are expected, considering

that when the injection channels have a smaller cross-sectional area, plugs of the dispersed fluid

with a smaller volume are periodically generated at the cross-junction through pinch-off. As the

injection channel expands into the larger test channel, the plugs in the injection channel changed

to droplets. At higher volumetric flow ratios the frequency of the plugs increases and so does the

102 injection of droplets into the larger test channel. The flow regime generated in the injector channel dictated the nature of the regime in the test channel.

5.4 Flow Maps

Based on the observed topological segmented flow regimes, Figure 5.12 and Figure 5.13 show the flow maps and transition lines between regimes from the Type I and II chips, respectively. The points in the figures represent each liquid-liquid segmented flow state with respect to the carrier fluid superficial velocity (JC) and the dispersed fluid superficial velocity

(JD). The superficial velocity is defined as the volumetric flow rate of the carrier or dispersed fluid divided by the cross sectional area of the test channel. The Droplet and Plug flow regimes were shifted toward the higher dispersed and lower carrier fluid superficial velocities with increasing expansion ratio. The Plug flow regime was broader with the lower expansion ratio channels. The transient Irregular Segmented flow was favored in the higher expansion ratio channel and the interval of transient Irregular Segmented flow between the Droplet and Plug flow regimes was shorter for the low expansion ratio channels. In the lowest expansion ratio chip, the Type III chip with an expansion ratio of 2, most of the flow patterns were regular Plug flows without transient irregular flow patterns.

The length of the dispersed fluid phases, including droplets and plugs, and the distance between the dispersed fluid phases were acquired by image processing and plotted as a function of the carrier fluid volumetric flow ratio in Figure 5.14. More than 100 image frames from the test channel of the Type II chip were recorded and used for image processing to ensure statistical integrity. The average length of the dispersed fluid scaled approximately with βC to the -1.2 power and the error bars represents the standard deviation of the longitudinal length scale of each phase [see Figure 5.14]. The broader error bars of the dispersed fluid length at carrier fluid

103

104

(a)

(b)

Figure 5.14 Measurement of the dispersed and carrier fluid lengths in terms of carrier fluid volumetric flow ratio from Type II chip.

105 volumetric flow ratios between 0.6 and 0.83 correspond to the Irregular Segmented flow regime.

Variation of the carrier fluid lengths are much larger than those of the dispersed fluid lengths for carrier fluid volumetric flow ratios larger than 0.6. As explained in previous work on gas-liquid two-phase flows [129], carrier fluids communicated with each other through the corner and potentially thin film flows between the dispersed fluid and channel walls. This caused much larger variation in the length of the carrier fluid plugs. The error bars of the carrier fluid lengths in the Irregular Segmented flow regimes (i.e., 0.6 < βC < 0.83) could be regarded as much longer because the contacting droplets or plugs without space between them were excluded from the data mining process, which possibly affected the trends of carrier fluid lengths with the carrier fluid volumetric flow ratio.

5.5 Wetting of Dispersed Fluid

One advantage of liquid-liquid segmented flow for bio-analytical applications is that each dispersed droplet or plug works as an independent biochemical reactor. Wetting of the channel surfaces with the dispersed fluid is not desirable since the adsorption of molecules on the channel walls would result in cross-contamination. Wetting of the channel walls with dispersed segmented flow was observed in these experiments. Once wetting nucleated on any microchannel surface, the flow regimes could no longer be predicted based on the carrier fluid volumetric flow ratio and the wetted patch on the surface caused unpredictable flows. This wetting commonly occurred after long plug flow experiments at low Capillary numbers; under these conditions longer aqueous plugs occurred due to breakdown of the carrier fluid liquid films.

Furthermore, the surface modification to make the channel walls more hydrophobic can be deteriorated after extended use, exposing wetting patches on the surface. Figure 5.15 shows examples of the dispersed fluid wetting the walls in the test channel in the form of droplets [see

106 Figure 5.15 (a)] and plugs [see Figure 5.15 (b)]. Wetting of the injection channel wall with the dispersed fluid induced different flow patterns even at the same carrier fluid volumetric flow ratio [see Figure 5.15 (c) and Figure 5.15 (d)]. Water deposited on the injection channel walls restrained following droplets until the drag force was high enough to cause release of the larger attached droplet. This resulted in Plug flow, rather than the Droplet flow generated by the non- wetted injection channel.

In order to obtain a consistent and predictable flow regime with a fixed volumetric flow ratio, wetting should be avoided. Pseudo-fluorination of a functionalized carboxylate polymer

(a) (b)

107 surface through UV irradiation would be a potential solution for the prevention of wetting by the

aqueous dispersed phase on the channel surface by making the channel surface more fluorophilic

and more hydrophobic. To return the wetted channel surface to its initial state, the same surface

treatment procedure outlined previously was repeated.

5.6 Flow Velocity Measurement

Measurement of the velocity of the dispersed droplets and plugs was accomplished by

recording two consecutive image frames separated by a known time interval ( ∆ ) and measuring the distance traveled (∆). A camera in double exposure mode and a 532 nm Nd:YAG double pulsed laser with an external timing controller (Solo II PIV, New Wave Research, Inc,

Fremont, CA) were used for the velocity measurement. Figure 5.16 (a) and (b) show the overlap of two consecutive images acquired with the double pulsed laser at a 2.5 msec pulse separation time in the Plug and Droplet flow regimes, respectively. The velocity of the dispersed fluid

was calculated by dividing the transit distance (∆) of a dispersed droplet or plug by the fixed pulse separation time (∆) as shown in Equation 5.3. Image processing was used to extract the

centroids of the droplets and plugs in each two-dimensional image frame and calculate the

distance traveled by the centroids during the pulse separation time. Measured velocities were

scaled by the sum of the superficial velocities of the dispersed and carrier fluids (Equation 5.4),

to show the difference between the superficial velocity and the measured velocity.

∆ (5.3) ∆

⁄ (5.4)

108

(a) (b)

(c)

109 The measured velocity was greater than the corresponding superficial velocity in some ranges

depending on the segmented flow regime. In the Droplet flow regime, which had no contact

between the droplets and the channel walls, the scale factor was 1.46 ± 0.077 with a coefficient

of variation of 5.3%. In the Plug flow regime, the scale factor was smaller than that in Droplet

flow at 1.25 ± 0.049 with a coefficient of variation 3.9% [see Figure 5.16 (c)].

There are two possible reasons for the decrease in the scale factor from the Droplet to the

Plug flow regimes. Smaller droplets centered within the channel are convected with a velocity closer to the maximum velocity of the velocity profile in the neighborhood of the centerline in

the micro-channel. Another factor is that the relatively larger volume of the plugs filled more of

the cross-section of the micro-channel and the higher drag force due to friction in the thin films

between the plugs and channel walls slowed them down.

5.7 Liquid-Liquid Segmented Flow Pressure Drop

While the micro-scale segmented flows can improve heat transfer, mixing efficiency,

analyte economy and isolation, there is an associated pressure drop increase with their use.

Prediction of the pressure drop is required prior to effectively implementing segmented flow for

practical applications in bio-analytical systems. The pressure drops associated with liquid-liquid segmented flows were measured in the Type II chip.

Figure 5.17 (a) illustrates the liquid-liquid segmented flow pressure drop in terms of the carrier fluid volumetric flow ratio. The pressure drop for each set of experiments was measured by increasing the volumetric flow rate of the dispersed fluid (QD) gradually while maintaining a

constant volumetric flow rate of the carrier fluid (QC). The pressure drop increased to a local

maximum with the dispersed fluid flow rate and fixed carrier fluid flow rate (i.e., lower carrier

fluid volumetric flow ratio) then decreased slightly at higher dispersed fluid flow rates.

110

(a)

(b)

Figure 5.17 (a) Liquid-liquid segmented flow pressure drops and (b) scaled segmented flow pressure drops by single liquid flow pressure drops as a function of the carrier fluid volumetric flow ratio.

111 Additional increases of the dispersed fluid flow rate generated longer dispersed fluid plugs and

increased the pressure drop again with higher characteristic velocity driven by the dispersed fluid

flow rate. Changes of the volumetric hold-up of the dispersed and carrier fluid for different

viscosities (i.e., viscosity ratio of dispersed and carrier fluid ≈ 1.87) across a given length affect

the pressure drop behavior with respect to the carrier volumetric flow ratio. Similar variation in

the two-phase flow pressure drop with the volumetric flow ratio was also observed from the

measurement of air-water two-phase flow pressure drop [129], which had greater pressure drops

with different volumetric hold-up of air and water in a fixed length of channel due to much

higher viscosity ratio between the gas and liquid (i.e., viscosity ratio of air and water ≈ 55).

Figure 5.17 (b) illustrates scaled liquid-liquid segmented flow pressure drops by single phase flow pressure drops, which were measured from the single fluorocarbon carrier fluid flows.

Single phase flow pressure drops for the carrier fluid were measured on the Type II chip under the same experimental conditions and used as a reference pressure for scaling the two- phase segmented flow pressure drop. The experimentally determined friction factor (f) for the single phase pressure drop using the Darcy-Weibach equation (Equation 5.5) is shown in Figure

5.18 as a function of Reynolds number (Re).

∆ 2 (5.5) ∆

where Δ⁄Δ is the frictional pressure drop per unit length, is the hydraulic diameter of microchannel, is the density of fluid, and is the characteristic velocity. The theoretical correlation of the friction factor and Reynolds number for a fully developed laminar flow is well- known to be 64⁄ . However, in previous experimental investigations [130, 131], the resulting constants varied from 50 to 104 depending on the experimental instrumentation and

112

Figure 5.18 Experimental measurement of friction factor, f, in terms of Reynolds number for the carrier fluid single phase flow pressure drop.

procedures. This difference may arise from several uncertainties, such as the variation of the channel geometry including the cross-section change along the channels and the serpentine configuration; the surface properties including the surface energy (wettability) due to surface

treatments and surface roughness; fluid property differences including density, and viscosity of the fluids used in the experiments. For the experiments using the Type II chip, a constant of 86.7 with a 19% coefficient of variation was determined experimentally and the Darcy-Weibach equation for the carrier fluid single phase pressure drop with the constant is shown in Equation

5.6.

∆ 43.35 (5.6) ∆

113 Using the Lockhart-Martinelli correlation, the two-phase flow pressure drop was predicted by multiplying the two-phase friction multiplier with the single phase flow pressure drop.

∆ ∆ (5.7) ∆ ∆

The subscripts TP and S denote two-phase and single phase flows respectively. The multiplier is correlated with the Lockhart-Martinelli parameter (X) as;

∆ ∆ (5.8) ∆ ∆

The relation between the friction multiplier and the parameter (X) is an inverse second-

order polynomial [119].

1 1 (5.9)

where C is a coefficient depending on flow conditions of dispersed and carrier fluids. The

measured segmented flow pressure drop data were plotted against the Lockhart-Martinelli parameter in Figure 5.19 (a). A coefficient, C, of 4.63 gave a good fit to the experimental data.

The measured liquid-liquid segmented flow pressure drop was compared with the predicted

pressure drops showed agreement within ± 20% as plotted in Figure 5.19 (b).

5.8 Conclusions

Liquid-liquid segmented flows in polymer microchannels were studied experimentally using polymer chips with three different expansion ratios (1:16, 1:4, and 1:2) from the injection

114

(a)

(b)

Figure 5.19 Measured two-phase friction multiplier data in terms of Lockhart-Martinelli parameter (b) comparison of measured and predicted liquid-liquid segmented flow pressure drop with C=4.63.

115 to test channels. Flow regimes and maps were determined based on the flow topology from each

chip. Effects of the different expansion ratios on the regimes and irregularity of segmented flows

were examined. The droplet and plug regimes were shifted to the higher carrier and lower

dispersed fluid superficial velocities and the plug flow regime was broader with the lower expansion ratio channels. The transient Irregular Segmented flow was favored in the higher expansion ratio channel and the transition interval between the Droplet and Plug flow regimes

was shorter for the low expansion ratio channels. Lengths of dispersed and carrier fluid were

measured in terms of the carrier fluid flow ratio and the effects of wetting of dispersed fluid at

the channel wall surface on flow regimes were discussed. Measured velocities of dispersed fluid

droplets and plugs were faster than the sum of the dispersed and liquid superficial velocities.

Liquid-liquid segmented flow pressure drops were measured and compared with the Lockhart-

Martinelli predictions.

116 Chapter 6 Conclusions and Future Work

6.1 Conclusions

6.1.1 Microfabrication of Polymer Chips

Three different types of microfluidic chips made from two thermoplastics, polycarbonate

(PC) and poly-methyl methacrylate (PMMA), were fabricated using direct micromachining of

the polymer, and hot embossing with micromilled brass and X-ray LIGA mold inserts. Direct

micromachining of the polymer provided an appropriate method for rapid prototyping of

microfluidic chips. Hot embossing of thermoplastic with brass and LIGA mold inserts gives a

way to produce high volumes of polymer microfluidic chips. Each mold insert had different

characteristics. The LIGA process enabled fabrication of polymer chips with delicate geometry

and high quality surfaces. While micromachining of the brass mold insert has limitations in

fabricating sharp corners and smooth surfaces, it produced reliable mold inserts in a quick

process time frame.

6.1.2 Experimental Study of Gas-Liquid Two-Phase Flow in Microchannels

Air-water two-phase flows in microchannels fabricated from poly-methyl methacrylate

(PMMA), with walls that were partially non-wetting, with a typical static contact angle 65° in stock form, and not molecularly smooth, were experimentally investigated. Two different types of chips were prepared: microchannels with unity aspect ratio replicated using hot embossing of

PMMA with a micro-milled brass mold insert and micro-milled microchannels of PMMA with aspect ratios of 1, 2 and 3 and a fixed hydraulic diameter.

Flow maps were obtained using the same gas-liquid injection geometry and methods for the three different aspect ratio microchannels, and the resulting regime boundaries compared to those found by other investigators. The bubbly flow regime boundary was shifted to higher

117 liquid and/or lower gas superficial velocities for the higher aspect ratio channels, while transition

to the Annular and Annular-Dry regimes remained the same, within experimental uncertainty.

Three sub-regimes of the Segmented flow regime were identified on the basis of the statistical

variation in the associated phase length scales from flow observations over a substantial channel length. Feedback effects were significant, but not the only cause of the segmented flow irregularities observed in the flows produced using the two different injection geometries and microchannel manufacturing techniques. Two other mechanisms were identified as responsible for the development of irregular Segmented flow under steady gas and liquid flow rates. One was related to the progressive mutual approach of bubble pairs along the channel resulting in coalescence and induced by small perturbations. This mechanism occurred mostly at relatively high liquid volumetric flow ratios. The other mechanism was attributed to unsteady injection because of the abrupt gas expansion and high liquid plug inertia at the transition from a smaller injection channel to the test channel with relatively low liquid volumetric flow ratios.

The variability of the size of the liquid plugs separating gas bubbles in Segmented flow was substantially higher than that of the bubbles, even when the flow was regular with low variability of the bubble size. Irregular Segmented flow was prevalent at higher liquid superficial velocities, lower liquid volumetric flow ratios, and lower channel aspect ratios. Of the three aspect ratios examined, the microchannels with an aspect-ratio of 3 displayed the broadest window of regular Segmented flow.

Measured two-phase flow pressure drops in the Bubbly, Segmented-Annular, Annular and Dry flow regimes, which were also observed in macro-scale channels agreed well with the pressure drops predicted using a Lockhart-Martinelli correlation with a C parameter of 1.39.

However, in the Segmented flow regimes, which are mainly observed in micro-scale channels,

118 the measured pressure drops were ~ 1.5 to 7 times higher than predicted. The Segmented flow

regime was subdivided into Segmented-1, -2, and -3 flow regimes based on topological

observations including the gas bubble and liquid plug lengths and the number of bubbles across

the microchannel. Each sub-regime was associated with different trends in the pressure drop

variation with respect to the liquid volumetric flow ratio.

Continuous flow of gas bubbles whose diameter was larger than the channel width was defined as Segmented flow. Therefore, each gas bubble occupied most of cross-section of the microchannel leaving a very thin liquid film between the gas bubble and channel wall and work as hydraulic resistance, in addition to the frictional resistance induced by channel wall. The

Segmented-1 flow regime, with the highest number of bubbles in regular and stable pattern across the microchannel, showed a linear increase in pressure drop due to the pressure build-up within the liquid plug blocked by the gas bubble. Another contribution to the pressure increase was the thin film subjected to the superficial velocity, which reduced for the liquid to flow through the film. The rising pressure in the Segmented-1 flow regime was scaled down and plateaued in the Segmented-2 flow regime because of the reduced number of gas bubbles and thickening of the liquid thin film as a result of the increased gas superficial velocity. The thickness of the liquid thin film is proportional to the Capillary number (i.e., flow velocity). The pressure built in each liquid plug blocked by a gas bubble was released more and run down in the

Segmented-3 flow regime due to the increased liquid film thickness between the gas bubble and

the channel wall, which provided a wider path for the liquid to flow through and released the

pressure build-up. In this regime, number of gas bubble also kept decreasing with increased gas

superficial velocity. Large density and viscosity ratios of air and deionized water made this up

and down of pressure more significant.

119 6.1.3 Experimental Study of Liquid-Liquid Segmented Flow in Microchannels

Liquid-liquid segmented flows in microchannels fabricated on polymer test chips were

investigated experimentally. Polymer test chips were prepared using hot embossing of polycarbonate (PC) sheets with micro-milled brass mold inserts. Three different configurations of microchannels were prepared with the injection to test channel expansion ratio of 16 (50 μm

width × 50 μm depth to 200 μm width × 200 μm depth), 4 (50 μm width × 200 μm depth to 200

μm width × 200 μm depth), and 2 (100 μm width × 200 μm depth to 200 μm width × 200 μm

depth). Deionized water with blue food-coloring dye (1% v/v) was used as a dispersed fluid at

flow rates (QD) between 0.5 and 60 µl/min. The carrier fluid was perfluorocarbon (FC 3283) with a nonionic fluorous-soluble surfactant (Perfluorooctanol, 10% v/v) at flow rates (QC) between 3 and 25 µl/min. The two fluids were injected into the chips separately.

While the Droplet and Plug flows with the transient Irregular Segmented flows between

two flow regimes were mainly observed in the test channels of the expansion ratio 16 and 4 chips,

only the Plug flow was observed in the test channel of the expansion ratio 2 chip. With only the

expansion ratio 16 chip, scattered Droplet flow was observed in addition to regular Droplet flow.

Flow pattern maps and transitions between flow regimes were determined in terms of a fixed

homogeneous carrier fluid volumetric flow ratio (βC) to compare the effect of the expansion

ratios from the injection to the test channels. The droplet and plug regimes were shifted to higher

carrier and lower dispersed fluid superficial velocities and the plug flow regime was broader with

the lower expansion ratio channels. The transient Irregular Segmented flow was favored in the

higher expansion channel ratio and the interval of transient Irregular Segmented flow between the Droplet and Plug flow regimes were shorter for the low expansion channel ratios. This is

evidence that the flow regime maps in microchannels are not universal and depend on the

120 configuration of the micro-injection system. The length of the dispersed segmented flows and the

distance between consecutive droplets or plugs as a function of the carrier fluid volumetric flow

ratio (βC) were determined by image processing of frames acquired via CCD camera with bright

field illumination. The average length of the dispersed fluid plugs was shown to scale

approximately with βC to the -1.2 power. Velocities of the dispersed droplet and plug flows were

measured using double-pulsed laser illumination and were found to be 1.46 ± 0.077 and 1.25 ±

0.049 times faster than the superficial velocity of the segmented flow respectively. Pressure drop

measurements of liquid-liquid segmented flows were also carried out for all flow regimes and

associated trends were correlated with changes in flow topology. Comparisons of the

experimental pressure drop with the predictions for a modified Lockhart-Martinelli correlation

with parameter C = 4.63 showed agreement within ± 20 %.

6.2 Future Work

6.2.1 High Throughput Bioassay Using Droplets and FCCS

As a practical application of multiphase flow in high throughput screening, monitoring of

endonuclease-inhibitor activity within continuously segmented droplets is underway by using fluorescence cross-correlation spectroscopy (FCCS) set up by Wonbae Lee of Dr. Soper’s research group in the Department of Chemistry at LSU [see Figure 6.1]. Fluorescence correlation spectroscopy (FCS) is one of the detection methods, which was developed to monitor relatively low concentrated single molecules through a correlation analysis of fluctuation of the fluorescence intensity. Combining droplet flow in a transparent polymer chip with FCCS enables detection of enzyme activity at the single molecule level. Most protein molecules tend to adsorb non-specifically on any surface [132]. Reduced reagent using the miniaturized bio-analytical systems with high surface to volume ratio may result in significant loss of active enzymes during

121

Figure 6.1 Monitoring of enzyme activity in droplet for high throughput bioassay using fluorescence cross-correlation spectroscopy (FCCS).

screening process for this reason. The loss of enzymes decreases detection efficiency due to reduced number of the enzyme-substrate reactions. Using multiphase flow, all molecules of interest can be preserved in encapsulated droplets without any enzyme loss. Cross-correlation analysis of the fluctuations of fluorescent signals from two fluorophores in distinct excitation- emission wavelengths (Cy3: ~550 nm excitation and ~570 nm emission, IRdye800: ~778 nm excitation and ~806 nm emission) conjugated with double stranded DNA confined in segmented droplets along with the endonulease and inhibitor allow the functionality of endonuclease and inhibitor in each segmented droplet to be observed. Figure 6.2 illustrates the instrument setup for fluorescence cross-correlation spectroscopy (FCCS) built by Wonbae Lee of the Department of

Chemistry at LSU. Two diode laser beams with 532nm and 780nm wavelengths were designed for the Cy3 and IRdye800, respectively. Emitted photons from the fluorophores excited by the

122

Figure 6.2 Schematic illustration of the fluorescence cross-correlation spectroscopy (FCCS) setup using dual laser sources, 532nm and 780 nm diode lasers, (Image of courtesy of Wonbae Lee – Department of Chemistry, LSU).

corresponding laser beams are detected by avalanche photo diode (APD) in 80 MHz (12.5 ns resolution) counting rate. The two intensity signals are cross-correlated using an in-house developed correlation program. As a preliminary experiment, detection of single signals from fluorescent polystyrene microspheres (FluoSpheres® F8809, 540/560, Invitrogen Corp.,

Carlsbad, CA) suspended in each droplet are shown in Figure 6.3 with respect to different carrier fluid volumetric flow ratios. Periodic peaks in the figures were acquired from each droplet flowing in a microchannel hot embossed in a polycarbonate chip with a 50 µm × 50 µm injection channel and a 200 µm × 200 µm test channel at different flow rates. Maintaining the volumetric flow rate of the carrier fluid at 20 μl/min, the volumetric flow rate of the dispersed fluid was increased gradually. The diameter of each droplet was around 100 μm (~ 500 picoliter).

123 Dispersed fluid

Carrier fluid

Detection point

(a)

(b) QD = 1 µl/min, QC = 20 µl/min (c) QD = 1 µl/min, QC = 20 µl/min

(d) QD = 1.8 µl/min, QC = 20 µl/min (e) QD = 2 µl/min, QC = 20 µl/min

124 Microscopic images of the droplet flow and fluorescent signal peaks shows increased frequency with increased dispersed fluid flow rate.

6.2.2 Measurement of Liquid Thin Film

Another meaningful work using FCCS is the measurement of the liquid thin film between the dispersed aqueous fluid and the microchannel wall. Distinct biochemical substances are required in the continuously segmented plugs for the high throughput screening. In order to maintain the original biochemical entity in each segmented plug, each plug needs to be separated from channel surface. Otherwise, traces of biomolecule from preceding plugs adsorb on the channel surface and may cause cross-contamination of the following plugs containing different entities. For this reason, survival of liquid thin film between the plugs and the channel walls is required under any fluidic condition.

To date, the thickness of the liquid thin films was estimated theoretically and computationally due to the lack of suitable detection methods for very thin liquid layers.

Detection of two fluorophores with different excitation and emission spectra in the carrier and dispersed fluids may allow measurement of the thickness of the thin film. Fluorescent polystyrene microspheres with 0.2 µm diameter (FluoSpheres® F8809, ~540 nm excitation and

~560 nm emission, Invitrogen Corp., Carlsbad, CA) and quantum dots (Qdots® ITK 800 organic quantum dots, broad band excitation and ~800 nm emission, Invitrogen Corp., Carlsbad, CA) with lipophilic surfaces could be used for aqueous dispersed fluid and organic carrier fluids, respectively. Figure 6.4 demonstrates the experimental scheme for the measurement of the liquid thin film between the dispersed liquid plug and the channel wall using fluorescence cross- correlation spectroscopy. Excitation and emission spectra of the fluorescent microspheres and quantum dots were spaced enough to increase the signal to noise ratio without interference

125 between the two emitted fluorescent signals. Additionally, translational driver with high spatial

resolution is necessary to scan the microchannel through the fixed laser beam. Stepper motor (~

20 nm resolution) and piezoelectric motor (~ 5 nm resolution) stages are good candidates to achieve the required high scanning resolution.

(a)

(b)

126 References

[1] P. S. Dittrich and A. Manz, "Lab-on-a-chip: microfluidics in drug discovery," Nature Reviews Drug Discovery, vol. 5, pp. 210-218, 2006.

[2] "ANSI/SBS 1, 2, 3, 4 - 2004: Microplates," American National Standard Institute/Society for Biomolecular Sciences, 2004.

[3] S.-i. Nihira, H. Kotaki, and A. Nakano, "High-throughput screening through the use of robotics and a high-performance data handling system," American Laboratory, vol. 29, pp. 15-18, 1997.

[4] H. F. Zimmermann and J. Rieth, "A Fully Automated Robotic System for High Throughput Fermentation," Journal of the Association for Laboratory Automation, vol. 11, pp. 134-137, 2006.

[5] K. R. Oldenburg, "Development of an ultra-high throughput screening system: Plate design, liquid handling, and image analysis," in Proceedings of SPIE, San Jose, CA, United States, 1998, pp. 197-208.

[6] D. A. Dunn and I. Feygin, "Challenges and solutions to ultra-high-throughput screening assay miniaturization: Submicroliter fluid handling," Drug Discovery Today, vol. 5, pp. S84-S91, 2000.

[7] J. Wolcke and D. Ullmann, "Miniaturized HTS technologies - uHTS," Drug Discovery Today, vol. 6, pp. 637-646, 2001.

[8] I. Schneegaß and J. M. Köhler, "Flow-through polymerase chain reactions in chip thermocyclers," Reviews in Molecular Biotechnology, vol. 82, pp. 101-121, 2001.

[9] M. Curcio and J. Roeraade, "Continuous segmented-flow polymerase chain reaction for high-throughput miniaturized DNA amplification," Analytical Chemistry, vol. 75, pp. 1-7, Jan 1, 2003.

[10] D. L. Chen and R. F. Ismagilov, "Microfluidic cartridges preloaded with nanoliter plugs of reagents: an alternative to 96-well plates for screening," Current Opinion in Chemical Biology, vol. 10, pp. 226-231, 2006.

[11] V. Linder, S. K. Sia, and G. M. Whitesides, "Reagent-Loaded Cartridges for Valveless and Automated Fluid Delivery in Microfluidic Devices," Analytical Chemistry, vol. 77, pp. 64-71, January 1, 2005.

[12] D. N. Adamson, D. Mustafi, J. X. J. Zhang, B. Zheng, and R. F. Ismagilov, "Production of arrays of chemically distinct nanolitre plugs via repeated splitting in microfluidic devices," Lab on a Chip, vol. 6, pp. 1178-1186, 2006.

127 [13] C. G. J. Schabmueller, A. G. R. Evans, A. Brunnschweiler, G. Ensell, D. L. Leslie, and M. A. Lee, "Design, fabrication and packaging of closed chamber PCR-chips for DNA amplification," in Proceedings of SPIE - The International Society for Optical Engineering, Paris, Fr, 2000, pp. 362-369.

[14] A. E. Guber, M. Heckele, D. Herrmann, A. Muslija, V. Saile, L. Eichhorn, T. Gietzelt, W. Hoffmann, P. C. Hauser, and J. Tanyanyiwa, "Microfluidic lab-on-a-chip systems based on polymers--fabrication and application," Chemical Engineering Journal, vol. 101, pp. 447-453, 2004.

[15] E. T. Lagally and R. A. Mathies, "Integrated genetic analysis microsystems," Journal of Physics D: Applied Physics, vol. 37, pp. 245-261, 2004.

[16] J. El-Ali, S. Gaudet, A. Gunther, P. K. Sorger, and K. F. Jensen, "Cell stimulus and lysis in a microfluidic device with segmented gas-liquid flow," Analytical Chemistry, vol. 77, pp. 3629-36, Jun 1, 2005.

[17] S. Wälchli and P. R. Von Rohr, "Two-phase flow characteristics in gas-liquid microreactors," International Journal of Multiphase Flow, vol. 32, pp. 791-806, 2006.

[18] P. Garstecki, M. J. Fuerstman, M. A. Fischbach, S. K. Sia, and G. M. Whitesides, "Mixing with bubbles: a practical technology for use with portable microfluidic devices," Lab on a Chip, vol. 6, pp. 207-12, Feb, 2006.

[19] T. Cubaud and C.-M. Ho, "Transport of bubbles in square microchannels," Physics of Fluids, vol. 16, pp. 4575-4585, 2004.

[20] T. Cubaud, U. Ulmanella, and C.-M. Ho, "Two-phase flow in microchannels with surface modifications," Fluid Dynamics Research, vol. 38, pp. 772-786, 2006.

[21] A. Günther, S. A. Khan, M. Thalmann, F. Trachsel, and K. F. Jensen, "Transport and reaction in microscale segmented gas-liquid flow," Lab on a Chip, vol. 4, pp. 278-286, 2004.

[22] A. Günther, M. Jhunjhunwala, M. Thalmann, M. A. Schmidt, and K. F. Jensen, "Micromixing of miscible liquids in segmented gas-liquid flow," Langmuir, vol. 21, pp. 1547-55, Feb 15, 2005.

[23] D. R. Link, S. I. Anna, D. A. Weitz, and H. A. Stone, "Geometrically Mediated Breakup of Drops in Microfluidic Devices," Physical Review Letters, vol. 92, pp. 54503-1, 2004.

[24] A. Kawahara, P. M. Y. Chung, and M. Kawaji, "Investigation of two-phase flow pattern, void fraction and pressure drop in a microchannel," International Journal of Multiphase Flow, vol. 28, pp. 1411-1435, 2002.

[25] N.-T. Nguyen and S. T. Wereley, Fundamentals and Applications of Microfluidics. Norwood, MA: Artech House, 2002.

128 [26] T. D. Perroud, J. N. Kaiser, J. C. Sy, T. W. Lane, C. S. Branda, A. K. Singh, and K. D. Patel, "Microfluidic-Based Cell Sorting of Francisella tularensis Infected Macrophages Using Optical Forces," Analytical Chemistry, vol. 80, pp. 6365-6372, 2008.

[27] P. S. Williams, M. Zborowski, and J. J. Chalmers, "Flow Rate Optimization for the Quadrupole Magnetic Cell Sorter," Analytical Chemistry, vol. 71, pp. 3799-3807, 1999.

[28] M. Yamada and M. Seki, "Microfluidic Particle Sorter Employing Flow Splitting and Recombining," Analytical Chemistry, vol. 78, pp. 1357-1362, 2006.

[29] D. Huh, J. H. Bahng, Y. Ling, H.-H. Wei, O. D. Kripfgans, J. B. Fowlkes, J. B. Grotberg, and S. Takayama, "Gravity-Driven Microfluidic Particle Sorting Device with Hydrodynamic Separation Amplification," Analytical Chemistry, vol. 79, pp. 1369-1376, 2007.

[30] D. J. Harrison, K. Fluri, K. Seiler, Z. Fan, C. S. Effenhauser, and A. Manz, "Micromachining a Miniaturized Capillary Electrophoresis-Based Chemical Analysis System on a Chip," Science, vol. 261, pp. 895-897, August 13, 1993, 1993.

[31] H.-P. Chou, M. A. Unger, and S. R. Quake, "A Microfabricated Rotary Pump," Biomedical Microdevices, vol. 3, pp. 323 -330, 2001.

[32] V. Hessel, H. Löwe, and F. Schönfeld, "Micromixers--a review on passive and active mixing principles," Chemical Engineering Science, vol. 60, pp. 2479-2501, 2005.

[33] N.-T. Nguyen and Z. Wu, "Micromixers—a review," Journal of Micromechanics and Microengineering, vol. 15, p. R1, 2005.

[34] F. F. Keil, Modeling of Process Intensification. Weinheim: Wiley-VCH, 2007.

[35] M. Muradoglu, A. Günther, and H. A. Stone, "A computational study of axial dispersion in segmented gas-liquid flow," Physics of Fluids, vol. 19, p. 072109, 2007.

[36] G. I. Taylor, "Deposition of a viscous fluids on the wall of a tube," Journal of Fluid Mechanics, vol. 10, pp. 161-165, 1961.

[37] E. J. Walsh, C. King, R. Grimes, and A. Gonzalez, "Influence of segmenting fluids on efficiency, crossing point and fluorescence level in real time quantitative PCR," Biomedical Microdevices, vol. V8, pp. 59-64, 2006.

[38] F. P. Bretherton, "The motion of long bubbles in tubes," Journal of Fluid Mechanics, vol. 10, pp. 166-188, 1961.

[39] M. T. Kreutzer, F. Kapteijn, J. A. Moulijn, and J. J. Heiszwolf, "Multiphase monolith reactors: Chemical reaction engineering of segmented flow in microchannels," Chemical Engineering Science, vol. 60, pp. 5895-5916, 2005.

129 [40] P. Aussillous and D. Quere, "Quick deposition of a fluid on the wall of a tube," Physics of Fluids, vol. 12, pp. 2367-2371, 2000.

[41] W. B. Kolb and R. L. Cerro, "Coating the inside of a capillary of square cross section," Chemical Engineering Science, vol. 46, pp. 2181-2195, 1991.

[42] T. C. Thulasidas, M. A. Abraham, and R. L. Cerro, "Bubble-train flow in capillaries of circular and square cross section," Chemical Engineering Science, vol. 50, pp. 183-199, 1995.

[43] A. L. Hazel and M. Heil, "The steady propagation of a semi-infinite bubble into a tube of elliptical or rectangular cross-section," Journal of Fluid Mechanics, vol. 470, pp. 91-114, 2002.

[44] T. S. Zhao and Q. C. Bi, "Co-current air-water two-phase flow patterns in vertical triangular microchannels," International Journal of Multiphase Flow, vol. 27, pp. 765- 782, 2001.

[45] J. W. Coleman and S. Garimella, "Characterization of two-phase flow patterns in small diameter round and rectangular tubes," International Journal of Heat and Mass Transfer, vol. 42, pp. 2869-2881, 1999.

[46] J. M. Mandhane, G. A. Gregory, and K. Aziz, "A flow pattern map for gas--liquid flow in horizontal pipes," International Journal of Multiphase Flow, vol. 1, pp. 537-553, 1974.

[47] J. Weisman, D. Duncan, J. Gibson, and T. Crawford, "Effects of fluid properties and pipe diameter on two-phase flow patterns in horizontal lines," International Journal of Multiphase Flow, vol. 5, pp. 437-462, 1979.

[48] R. W. Lockhart and R. C. Martinelli, "Proposed correlation of data for isothermal two- phase, two-component flow in pipes.," Chem. Eng. Progress, vol. 45, pp. 39-48, 1949.

[49] H. J. Lee and S. Y. Lee, "Pressure drop correlations for two-phase flow within horizontal rectangular channels with small heights," International Journal of Multiphase Flow, vol. 27, pp. 783-796, 2001.

[50] T. Thorsen, R. W. Roberts, F. H. Arnold, and S. R. Quake, "Dynamic pattern formation in a vesicle-generating microfluidic device," Physical Review Letters, vol. 86, pp. 4163- 4166, 2001.

[51] H. Song, J. D. Tice, and R. F. Ismagilov, "A microfluidic system for controlling reaction networks in time," Angewandte Chemie - International Edition, vol. 42, pp. 768-772, 2003.

[52] L. Li, D. Mustafi, Q. Fu, V. Tereshko, D. L. Chen, J. D. Tice, and R. F. Ismagilov, "Nanoliter microfluidic hybrid method for simultaneous screening and optimization validated with crystallization of membrane proteins," Proceedings of the National Academy of Sciences, vol. 103, pp. 19243-19248, December 19, 2006, 2006.

130 [53] S. Mohr, Y. H. Zhang, A. Macaskill, P. J. R. Day, R. W. Barber, N. J. Goddard, D. R. Emerson, and P. R. Fielden, "Numerical and experimental study of a droplet-based PCR chip," Microfluidics and Nanofluidics, vol. 3, pp. 611-621, 2007.

[54] N. R. Beer, E. K. Wheeler, L. Lee-Houghton, N. Watkins, S. Nasarabadi, N. Hebert, P. Leung, D. W. Arnold, C. G. Bailey, and B. W. Colston, "On-Chip Single-Copy Real- Time Reverse-Transcription PCR in Isolated Picoliter Droplets," Analytical Chemistry, vol. 80, pp. 1854-1858, 2008.

[55] A. Gonzalez, D. Ciobanu, M. Sayers, N. Sirr, T. Dalton, and M. Davies, "Gene transcript amplification from cell lysates in continuous-flow microfluidic devices," Biomedical Microdevices, vol. 9, pp. 729-736, 2007.

[56] M. M. Kiss, L. Ortoleva-Donnelly, N. R. Beer, J. Warner, C. G. Bailey, B. W. Colston, J. M. Rothberg, D. R. Link, and J. H. Leamon, "High-Throughput Quantitative Polymerase Chain Reaction in Picoliter Droplets," Analytical Chemistry, vol. 80, pp. 8975-8981, 2008.

[57] A. S. Utada, E. Lorenceau, D. R. Link, P. D. Kaplan, H. A. Stone, and D. A. Weitz, "Monodisperse double emulsions generated from a microcapillary device," Science, vol. 308, pp. 537-541, 2005.

[58] T. Nisisako, S. Okushima, and T. Torii, "Controlled formulation of monodisperse double emulsions in a multiple-phase microfluidic system," Journal of Materials Chemistry, vol. 15, pp. 23-27, 2005.

[59] G. T. Kovacs, Micromachined Transducers Sourcebook. New York, NY: McGraw-Hill, 1998.

[60] M. J. Madou, Fundamentals of Microfabrication : The Science of Miniaturization, 2nd ed. Boca Raton, FL: CRC Press, 2002.

[61] B. M. Paegel, C. A. Emrich, G. J. Wedemayer, J. R. Scherer, and R. A. Mathies, "High throughput DNA sequencing with a microfabricated 96-lane capillary array electrophoresis bioprocessor," in Proc. Natl. Acad. Sci., 2002, pp. 574-579.

[62] P. C. Simpson, D. Roach, A. T. Woolley, and T. Thorsen, "High-throughput genetic analysis using microfabricated 96-sample capillary array electrophoresis microplates," Proceedings of the National Academy of Sciences of the United States of America, vol. 95, p. 2256, 1998.

[63] J. S. Marcus, W. F. Anderson, and S. R. Quake, "Parallel picoliter RT-PCR assays using microfluidics," Analytical Chemistry, vol. 78, pp. 956-958, 2006.

[64] J. W. Hong, V. Studer, G. Hang, W. F. Anderson, and S. R. Quake, "A nanoliter-scale nucleic acid processor with parallel architecture," Nature Biotechnology, vol. 22, pp. 435- 439, 2004.

131 [65] B. G. De Geest, J. P. Urbanski, T. Thorsen, J. Demeester, and S. C. De Smedt, "Synthesis of monodisperse biodegradable microgels in microfluidic devices," Langmuir, vol. 21, pp. 10275-10279, 2005.

[66] K. Ahn, C. Kerbage, T. P. Hunt, R. M. Westervelt, D. R. Link, and D. A. Weitz, "Dielectrophoretic manipulation of drops for high-speed microfluidic sorting devices," Applied Physics Letters, vol. 88, pp. 024104-3, 2006.

[67] J. D. Tice, H. Song, A. D. Lyon, and R. F. Ismagilov, "Formation of Droplets and Mixing in Multiphase Microfluidics at Low Values of the Reynolds and the Capillary Numbers," Langmuir, vol. 19, pp. 9127-9133, October 28, , 2003.

[68] G. K. M. Chantal, "Laser processing for bio-microfluidics applications (part I)," Analytical and Bioanalytical Chemistry, vol. V385, pp. 1351-1361, 2006.

[69] G. K. M. Chantal, "Laser processing for bio-microfluidics applications (part II)," Analytical and Bioanalytical Chemistry, vol. V385, pp. 1362-1369, 2006.

[70] W. Pfleging, A. Meier, T. Hanemann, H. Gruhn, and K. H. Zum Gahr, "Laser micromachining of metallic mold inserts for replication techniques," San Francisco, CA, 2000, pp. 551-556.

[71] T. Hanemann, W. Pfleging, J. Hausselt, and K. H. Zum Gahr, "Laser micromaching and light induced reaction injection molding as suitable process sequence for the rapid fabrication of microcomponents," Microsystem Technologies, vol. 7, pp. 209-214, 2002.

[72] D. Snakenborg, G. Perozziello, H. Klank, O. Geschke, and J. P. Kutter, "Direct milling and casting of polymer-based optical waveguides for improved transparency in the visible range," Journal of Micromechanics and Microengineering, vol. 16, pp. 375-381, 2006.

[73] M. L. Hupert, W. Jason Guy, S. D. Llopis, C. Situma, S. Rani, D. E. Nikitopoulos, and S. A. Soper, "High-precision micromilling for low-cost fabrication of metal mold masters," in Proceedings of SPIE, San Jose, CA, United States, 2006, p. 61120.

[74] L. Yi, W. Xiaodong, L. Chong, L. Zhifeng, C. Denan, and Y. Dehui, "Swelling of SU-8 structure in Ni mold fabrication by UV-LIGA technique," Microsystem Technologies, vol. 11, pp. 1272-1275, 2005.

[75] M. Hashimoto, P.-C. Chen, M. W. Mitchell, D. E. Nikitopoulos, S. A. Soper, and M. C. Murphy, "Rapid PCR in a continuous flow device," Lab on a Chip, vol. 4, pp. 638-645, 2004.

[76] M. Matteucci, F. Perennes, B. Marmiroli, P. Miotti, L. Vaccari, A. Gosparini, A. Turchet, and E. Di Fabrizio, "Compact micropumping system based on LIGA fabricated microparts," Microelectronic Engineering, vol. 83, pp. 1288-1290, 2006.

132 [77] C. K. Malek and V. Saile, "Applications of LIGA technology to precision manufacturing of high-aspect-ratio micro-components and -systems: a review," Microelectronics Journal, vol. 35, pp. 131-143, 2004.

[78] D. Gibbs and T. M. Crandell, An Introduction to CNC Machining and Programming. New York, NY: Industrial Press, 1991.

[79] P. Smid, CNC Programming Handbook, 2nd ed. New York, NY: Industrial Press, 2002.

[80] J. Chen, M. Wabuyele, H. Chen, D. Patterson, M. Hupert, H. Shadpour, D. Nikitopoulos, and S. A. Soper, "Electrokinetically synchronized polymerase chain reaction microchip fabricated in polycarbonate," Analytical Chemistry, vol. 77, pp. 658-666, 2005.

[81] Y. Li, J. S. Buch, F. Rosenberger, D. L. DeVoe, and C. S. Lee, "Integration of Isoelectric Focusing with Parallel Sodium Dodecyl Sulfate Gel Electrophoresis for Multidimensional Protein Separations in a Plastic Microfludic Network," Analytical Chemistry, vol. 76, pp. 742-748, 2004.

[82] E. M. Abad-Villar, J. Tanyanyiwa, M. T. Fernandez-Abedul, A. Costa-Garcia, and P. C. Hauser, "Detection of human immunoglobulin in microchip and conventional capillary electrophoresis with contactless conductivity measurements," Analytical Chemistry, vol. 76, pp. 1282-1288, 2004.

[83] M. Galloway, W. Stryjewski, A. Henry, S. M. Ford, S. Llopis, R. L. McCarley, and S. A. Soper, "Contact conductivity detection in poly(methyl methacylate)-based microfluidic devices for analysis of mono- and polyanionic molecules," Analytical Chemistry, vol. 74, pp. 2407-2415, 2002.

[84] H. Shadpour and S. A. Soper, "Two-Dimensional Electrophoretic Separation of Proteins Using Poly(methyl methacrylate) Microchips," Analytical Chemistry, vol. 78, pp. 3519- 3527, June 1, 2006, 2006.

[85] H. Shadpour, H. Musyimi, J. Chen, and S. A. Soper, "Physiochemical properties of various polymer substrates and their effects on microchip electrophoresis performance," Journal of Chromatography A, vol. 1111, pp. 238-251, 2006.

[86] D. Drechsler, A. Karbach, and H. Fuchs, "Nanoindentation on polycarbonate/polymethyl methacrylate Blends," Applied Physics A: Materials Science & Processing, vol. 66, pp. S825-S829, 1998.

[87] E. W. Backer, W. Ehrfeld, D. Münchmeyer, H. Betz, A. Heuberger, S. Pongratz, W. Glashauser, H. J. Michel, and R. Siemens, "Production of separation-nozzle systems for uranium enrichment by a combination of X-ray lithography and galvanoplastics," Naturwissenschaften, vol. 69, pp. 520-523, 1982.

[88] E. W. Backer, W. Ehrfeld, P. Hagmann, A. Maner, and D. Münchmeyer, "Fabrication of microstructures with high aspect ratios and great structural heights by synchrotron

133 radiation lithography, galvanoforming, and plastic moulding (LIGA process)," Microelectronic Engineering, vol. 4, pp. 35-56, 1986.

[89] E. Wolfgang and S. Andreas, "Recent developments in deep x-ray lithography," Journal of Vacuum Science & Technology B, vol. 16, pp. 3526-3534, 1998.

[90] W. Menz, J. Mohr, and O. Paul, Microsystem Technology, 2Rev ed. Weinheim, Germany: Wiley-VCH, 2001.

[91] C. Marques and K. W. Kelly, "Fabrication and performance of a pin fin micro heat exchanger," Journal of Heat Transfer, vol. 126, pp. 434-444, 2004.

[92] T. Wang, A. B. McCandless, R. M. Lienau, K. W. Kelly, D. Hensley, Y. Desta, and Z.-G. Ling, "Fabrication of monolithic multilevel high-aspect-ratio ferromagnetic devices," Journal of Microelectromechanical Systems, vol. 14, pp. 400-409, 2005.

[93] M. L. Hupert, M. A. Witek, Y. Wang, M. W. Mitchell, Y. Liu, Y. Bejat, D. E. Nikitopoulos, J. Goettert, M. C. Murphy, and S. A. Soper, "Polymer-based microfluidic devices for biomedical applications," in Proceedings of SPIE, San Jose, CA, United States, 2003, pp. 52-64.

[94] M. W. Mitchell, X. Liu, Y. Bejat, D. E. Nikitopoulos, S. A. Soper, and M. C. Murphy, "Modeling and validation of a molded polycarbonate continuous flow polymerase chain reaction device," San Jose, CA, United States, 2003, pp. 83-98.

[95] M. Heckele, W. Bacher, and K. D. Müller, "Hot embossing - The molding technique for plastic microstructures," Microsystem Technologies, vol. V4, pp. 122-124, 1998.

[96] H. Becker and U. Heim, "Hot embossing as a method for the fabrication of polymer high aspect ratio structures," Sensors and Actuators A: Physical, vol. 83, pp. 130-135, 2000.

[97] Y. H. Kim and R. P. Wool, "A theory of healing at a polymer-polymer interface," Macromolecules, vol. 16, pp. 1115-1120, 1983.

[98] R. P. Wool and K. M. O'Connor, "A theory of crack healing in polymers," Journal of Applied Physics, vol. 52, pp. 5953-5963, 1981.

[99] A. Aradian, E. Raphael, and P. G. de Gennes, "Strengthening of a Polymer Interface: Interdiffusion and Cross-Linking," Macromolecules, vol. 33, pp. 9444-9451, 2000.

[100] K. R. King, C. C. J. Wang, M. R. Kaazempur-Mofrad, J. P. Vacanti, and J. T. Borenstein, "Biodegradable Microfluidics," Advanced Materials, vol. 16, pp. 2007-2012, 2004.

[101] K.-S. Chen, A. A. Ayon, X. Zhang, and S. M. Spearing, "Effect of process parameters on the surface morphology and mechanical performance of silicon structures after deep reactive ion etching (DRIE)," Journal of Microelectromechanical Systems, vol. 11, pp. 264-275, 2002.

134 [102] A. Günther and K. F. Jensen, "Multiphase microfluidics: from flow characteristics to chemical and materials synthesis," Lab on a Chip, vol. 6, pp. 1487-1503, 2006.

[103] E. Evans, "Two-Phase Flow in High Aspect Ratio, Polymer Microchannels for Reagent Delivery Applications," M.S., Thesis, Mechanical Engineering, Louisiana State University, Baton Rouge, 2006.

[104] K. A. Triplett, S. M. Ghiaasiaan, S. I. Abdel-Khalik, and D. L. Sadowski, "Gas-liquid two-phase flow in microchannels Part I: two-phase flow patterns," International Journal of Multiphase Flow, vol. 25, pp. 377-394, 1999.

[105] G. Matsui, "Charateristic structure of upward bubble flow," in Dynamics of Two-phase Flows, O. C. Jones and I. Michiyoshi, Eds. New York: Begell House, 1992, pp. 359-377.

[106] V. E. Nakoryakov, O. N. Kashinsky, A. P. Burdukov, and V. P. Odnoral, "Local characteristics of upward gas-liquid flows," International Journal of Multiphase Flow, vol. 7, pp. 63-81, 1981.

[107] D. E. Nikitopoulos, J. Fiedler, M. Dowden, and E. Evans, "Three-dimensional bubble motion characteristics in vertical-upwards pipe flow," in International Conference on Multiphase Flows, Yokohama, Japan, 2004, p. 261.

[108] F. C. Knopf, J. M. Richard, G. Rice, and D. E. Nikitopoulos, "Pulsing to improve bubble column performance: I. Low gas rates," AIChE Journal, vol. 52, pp. 1103-1115, 2006.

[109] A. R. Berker, "Intétration des équations du movement d'un fluide visqueux incompressible," Encyclopedia of Physics, vol. 8, pp. 1-384, 1963.

[110] R. C. Rosenberg and D. C. Karnopp, Introduction to Physical System Dynamics. New York: McGraw-Hill, 1983.

[111] R. K. Shah and A. L. London, Laminar Flow Forced Convection in Ducts: A Source Book for Compact Heat Exchanger Analytical Data. New York: Academic Press, 1978.

[112] G. O. Brown, "The history of the Darcy-Weisbach equation for pipe flow resistance," Washington, DC, United States, 2002, pp. 34-43.

[113] W. H. McAdams, Heat Transmission, 3 rd ed. New York: McGraw-Hill, 1954.

[114] C. L. A. Cicchitti, M. Silvestri, G. Solddaini and R. Zavalluilli, "Two-phase cooling experiments––Pressure drop, heat transfer and burnout measurement," Energia Nucl, vol. 7, pp. 407-425, 1960.

[115] W. L. Owens, "Two-phase pressure gradient," in Int. Dev. in Heat Transfer. vol. Pt II New York: ASME, 1961.

[116] M. W. A.E. Dukler, III and R.G. Cleveland, "Pressure drop and hold-up in two-phase flow," ALChE J., vol. 10, pp. 38-51, 1964.

135 [117] D. R. H. Beattie and P. B. Whalley, "A simple two-phase frictional pressure drop calculation method," International Journal of Multiphase Flow, vol. 8, pp. 83-87, 1982.

[118] S. Lin, C. C. K. Kwok, R. Y. Li, Z. H. Chen, and Z. Y. Chen, "Local frictional pressure drop during vaporization of R-12 through capillary tubes," International Journal of Multiphase Flow, vol. 17, pp. 95-102, 1991.

[119] D. Chisholm, "A theoretical basis for the Lockhart-Martinelli correlation for two-phase flow," International Journal of Heat and Mass Transfer, vol. 10, pp. 1767-1778, 1967.

[120] K. Mishima and T. Hibiki, "Some characteristics of air-water two-phase flow in small diameter vertical tubes," International Journal of Multiphase Flow, vol. 22, pp. 703-712, 1996.

[121] H. Wong, C. J. Radke, and S. Morris, "The motion of long bubbles in polygonal capillaries. Part 2. Drag, fluid pressure and fluid flow," Journal of Fluid Mechanics, vol. 292, pp. 95-110, 1995.

[122] M. T. Kreutzer, F. Kapteijn, J. A. Moulijn, C. R. Kleijn, and J. J. Heiszwolf, "Inertial and interfacial effects on pressure drop of Taylor flow in capillaries," AIChE Journal, vol. 51, pp. 2428-2440, 2005.

[123] J. R. Burns and C. Ramshaw, "The intensification of rapid reactions in multiphase systems using slug flow in capillaries," Lab on a Chip, vol. 1, pp. 10-5, Sep, 2001.

[124] C. S. Jeanne, L. R. David, H. L. Thomas, P. L. Allen, H. P. Sapun, and A. F. Daniel, "Unilamellar vesicle formation and encapsulation by microfluidic jetting," Proceedings of the National Academy of Sciences of the United States of America, vol. 105, pp. 4697- 4702, 2008.

[125] D. K. Hwang, D. Dendukuri, and P. S. Doyle, "Microfluidic-based synthesis of non- spherical magnetic hydrogel microparticles," Lab on a Chip, vol. 8, pp. 1640-1647, 2008.

[126] M. Yamada, S. Doi, H. Maenaka, M. Yasuda, and M. Seki, "Hydrodynamic control of droplet division in bifurcating microchannel and its application to particle synthesis," Journal of Colloid and Interface Science, vol. 321, pp. 401-407, 2008.

[127] F. K. Hansen and G. Rødsrud, "Surface tension by pendant drop : I. A fast standard instrument using computer image analysis," Journal of Colloid and Interface Science, vol. 141, pp. 1-9, 1991.

[128] N. Kim, E. T. Evans, D. S. Park, D. E. Nikitopoulos, S. A. Soper, and M. C. Murphy, "Gas-liquid two-phase flow in hot embossed square microchannels," in ASME International Mechanical Engineering Congress and Exposition (IMECE2007), Seattle, WA, United States 2007, pp. 427-436.

136 [129] N. Kim, E. T. Evans, D. E. Nikitopoulos, S. A. Soper, and M. C. Murphy, "Investigation of two-phase flow in rectangular micro-channels," in ASME Fluids Engineering Conference (FEDSM2008), Jacksonville, FL, United States, 2008.

[130] I. Papautsky, J. Brazzle, T. Ameel, and A. B. Frazier, "Laminar fluid behavior in microchannels using micropolar fluid theory," Sensors and Actuators A: Physical, vol. 73, pp. 101-108, 1999.

[131] M. E. Steinke and S. G. Kandlikar, "Single-phase liquid friction factors in microchannels," International Journal of Thermal Sciences, vol. 45, pp. 1073-1083, 2006.

[132] A. R. Prakash, M. Amrein, and K. Kaler, "Characteristics and impact of Taq enzyme adsorption on surfaces in microfluidic devices," Microfluidics and Nanofluidics, vol. 4, pp. 295-305, 2008.

[133] MicroChem, "NANOTM SU-8 Negative Tone Photoresist Formulations 2-25," MicroChem Corp., Newton, MA.

[134] "MEGAPOSITTM SPRTM 220 SERIES PHOTORESISTS," Rohm and Hass Electronic Materials, Marlborough, MA 2004.

[135] Y. M. Desta, "Fabrication of High Aspect Ratio Vibrating Cylinder Microgyroscope Structures by Use of the LIGA Process," Ph.D., Dissertation, Mechanical Engineering, Louisiana State University, Baton Rouge, LA, 2005.

[136] C. Harris, Y. Desta, K. W. Kelly, and G. Calderon, "Inexpensive, quickly producable X- ray mask for LIGA," Microsystem Technologies, vol. V5, pp. 189-193, 1999.

[137] L. Jian, B. Loechel, H.-U. Scheunemann, M. Bednarzik, and A. Firsov, "Vitreous Carbon Membrane X-ray Masks for LIGA Process," in International Commercialization of Micro and Nano Systems Conference 2003 Amsterdam, The Netherlands, 2003.

[138] DisChem, "Technical Data Sheet - E-Liminate Pit Wetting Agent," DisChem Inc., Ridgway, PA.

[139] DisChem, "Technical Data Sheet - E-Form Sulfamate Nickel Concetrate," DisChem Inc., Ridgway, PA.

[140] DisChem, "Technical Data Sheet - DisChem Boric Acid," DisChem Inc., Ridgway, PA.

[141] DisChem, "Technical Data Sheet - E-Line pH Sulfamic Acid," DisChem Inc., Ridgway, PA.

137 Appendix A X-Ray LIGA Process

A.1 Optical and X-Ray Masks

The X-ray microfabrication begins with converting a design in CAD file format onto the optical mask which is used for UV lithography. This optical mask functioned as an intermediate mask for fabricating an X-ray mask. In this fabrication process, a 13 µm thick polyimide membrane (Kapton HN®, Goodfellow Corporation, Devon, PA) and a 4″ diameter, 225 um thick graphite disc (DFP-3, Poco Graphite, Inc., Decatur, TX) were selected as X-ray mask substrates.

The selection of photoresists (positive or negative) and optical mask patterns (clear field or dark field) depends on the physical properties of the two substrates. A negative photoresist, SU-8-10

(MicroChem Corp., Newton, MA) [133], was used on the graphite and needed a dark field optical mask (Figure A. 1 (b)). A positive photoresist, SPR 220-7 (Rohm and Haas Electronic

Materials, Marlborough, MA) [134], was used on the Kapton® membrane with a clear field mask (Figure A. 1 (a), Advanced Reproductions, North Andover, MA).

(a) (b)

Figure A. 1 (a) Clear field optical mask for positive photoresist. (b) Dark field optical mask for negative photoresist.

138 Graphite X-ray Mask

Graphite disks must be polished by lab tissue to remove carbon powder produced while slicing the graphite rod into disks. Graphite is a porous material, so dehydration is required before the spin coat of photoresist in convection oven at 110 °C for 15 minutes. SU-8 10 was

spin coated at 1,000 rpm for 20 seconds to obtain ~30 um thick of resist. In order to evaporate

solvent in the coated resist and solidify the film, a soft bake process was necessary before UV

exposure. A level hot plate in the cleanroom at CAMD was programmed for two different

temperature steps. The temperature was ramped from room temperature to 65 °C, maintained for

6 minutes, then heated to 95 °C and held for 14 minutes. After the pre-bake, the hot plate was

cooled down to room temperature.

UV light in the i-line (365 nm wavelength) from the UV station (Quintel UL7000-OBS

Aligner and DUV Exposure Station, Quintel Corp., Morgan Hill, CA) at CAMD irradiated the

SU-8 through the dark field optical mask (Figure A. 2(b)). Intensity of the UV light was

measured in range of 11.15 ~ 11.57 watts/cm2 during each exposure to calculate the appropriate

exposure time. A 450 mJ/cm2 dose for the 30 µm thick SU-8 10 at a UV intensity of 11.15

watt/cm2 resulted in 40 seconds exposure time. To intensify the cross-linking of SU-8 in the

exposed area, a post-exposure bake on a hot plate was followed. Temperature was ramped up to

65 °C, held for 2 minutes, ramped up to 95 °C again, maintained for 6 minutes, and cooled down

slowly to room temperature, completing the post-exposure bake. Post-exposure baked SU-8 was developed by immersion in SU-8 Developer (MicroChem Corp., Newton, MA) for 5 minutes with moderate agitation (Figure A. 2 (d)). After development, the substrate was rinsed with isopropyl alcohol (IPA) and dried by with nitrogen gas. The graphite substrate with patterned

SU-8 was glued on a NIST ring as shown in Figure A. 3 (a).

139 UV expose UV expose Optical mask Positive photoresist Negative photoresist (SJR 5740, (SU-8) SPR 220-7) Absorber: Chrome

Kapton® Membrane (a) (b) Graphite

Developed PR

Eletroplated gold

NIST ring

(e) (f)

Figure A. 2 Fabrication of Kapton® (Polyimide) and graphite X-ray masks: (a) UV exposure of positive photoresist (SJR 5740, SPR 220-7) on a Kapton® membrane; (b) UV exposure of negative photoresist (SU-8) on a graphite substrate; (c) Develop positive photoresist in UV exposed area; (d) Develop negative photoresist of UV unexposed areas; (e),(f) Gold electroplating for X-ray absorber.

Kapton® (Polyimide) X-ray Mask

A 13 um thin Kapton® (Polyimide) membrane was glued on the rim of a ring-shaped

stretcher and left overnight to give sufficient time for the glue to be cured. Following a good

bond, the Kapton® membrane was stretched radially. In order to release the internal stress built

by tension of the thin film, the stretched thin membrane was annealed in a convection oven at

90 °C for 20 minutes. After releasing the internal stress in the Kapton® membrane, it was

bonded to a NIST ring. 5 nm thickness of chrome (Cr) and 30 nm thickness of gold (Au) was

electron-beam deposited on the taut polyimide to make the polymer membrane conductive

substrate. These chrome and gold seed layers enabled electrodepostion of gold to be on the

nonconductive polyimide membrane. Following e-beam deposition, SPR 220-7 positive resist

140 was spin coated at 1,200 rpm for 30 seconds to reach ~15 um thick of resist and soft baked at

115 °C for 5 minutes in a convection oven. UV light 11.57 watts/cm2 i-line was exposed for 21.6 seconds to get 250 mj/cm2 exposure energy through the clear field optical mask shown in Figure

A. 2 (a). After post-exposure baking of the exposed resist in the convection oven at 115 °C for 2 hours, the exposed area on resist was developed by immersion in the M453 developer (Shipley,

Marlborough, MA) for 3 minutes with moderate agitation (Figure A. 2 (c)). Residue of the developed resist was rinsed by de-ionized water and dried with a gentle flow of nitrogen gas.

Table A. 1 summarizes process parameters in fabrication of the graphite and Kapton® X-ray masks. Figure A. 3 (c) shows a polyimide membrane with the patterned resist on the modified

NIST ring.

Table A. 1 Process parameters for fabrication of graphite and Kapton® X-ray masks.

Mask substrate Graphite disc Kapton® membrane

Optical mask pattern Clear field Dark field SU-8 10 SJR 5740, SPR 220-7 Photoresiste (Negative resist) (Positive resist) Spin coat 1,000 rpm for 20 sec, ~ 30um 1,200 rpm for 30 sec, ~15um 65 °C for 6 min 115 °C for 5 min Soft bake 95 °C for 14 min w/ hot plate in convection oven 365nm wavelength w/ intensity 365nm wavelength w/ intensity UV exposure 11.15 watt/cm2 11.57 watt/cm2 Dose 450 mj/cm2 250 mj/cm2

Exposure time ~ 40sec ~ 21.6 s 65 °C for 2 min 115 °C for 2 hours Post-exposure bake 95 °C for 6 min w/ hot plate in convection oven Developer SU-8 developer Shipley 354, M453

141 A.2 Gold Electroplating for X-Ray Absorber

Gold is typically used as an X-ray absorber in X-ray lithography. A sulfite-based

electrolyte, TECHNI-GOLD 25E (Technic, Inc., Cranston, RI) containing gold concentration in

the range of 8~16 g/L, was used to electrodeposit an 11 µm gold layer into the patterned

photoresist on the graphite and polyimide mask membranes. A 2 mA/cm2 current density from a

potentiostat (Model 2055, Amel Instruments, Milano, Italy) was applied over the opened area of

(a) (b)

Figure A. 3 Two different X-ray masks. (a) Patterned SU-8 on the graphite substrate. (b) Electroplated gold on the graphite. (c) Patterned SJR 5740 on the Kapton® membrane. (d) Electroplated gold on the Kapton®.

142

76 cm2 for 1.7 hours to get at least a 11 µm thick gold layer. A 7.5 um/hr deposit rate was

expected with 2 mA/cm2 applied [135]. A platinum coated titanium mesh anode (Technic, Inc.,

Cranston, RI) was used as a counter electrode. The temperature of the plating bath was

maintained at 55 °C with moderate stirring. Figures 3.9 (b) and (d) show graphite and polyimide

X-ray masks with completed gold absorber layer.

A.3 Preparation of Photo Resist and Stainless Steel Substrate

5″ diameter stainless steel plates were purchased from Mezzo Technologies (Baton

Rouge, LA) with surface flatness less than 5 µm. In order to improve adhesion between the

substrate and photoresist, a rough surface (Ra ≈ 600nm) was achieved by sand blasting. A 3mm

thick layer of monomer cast CQ grade PMMA sheet (Goodfellow Corporation, Devon, PA) of high molecular weight was prepared as a positive photoresist for X-ray lithography. Disks of 4″

diameter cut from the sheet were annealed and dehydrated in a convection oven at 80 °C for 1

hour (Figure A. 4 (a)).

(a) (b) (c)

Figure A. 4 Preparation of CQ grade PMMA photoresist and stainless steel substrate. (a) Before bonding. (b) Boding between PMMA disc and stainless steel substrate using PMMA adhesive. (c) Fly cutting of PMMA to the desired thickness.

143 Table A. 2 PMMA bonding solutions.

Bonding solution Weight

Benzoylperoxide (BPO) 0.15 g

N-Dimethylaniline (DMA) 0.1 g

3-Methacryloxypropyltrimethoxysilane (MEMO) 0.1 g

Methyl methacrylate Monomer (MMA) 85 g

Polymethyl methacrylate (PMMA) pellets 15 g

A PMMA adhesive was made by mixing chemicals listed in the Table A. 2. These

chemicals were mixed uniformly by vortexer and placed in vacuum chamber to eliminate

bubbles generated during mixing. Dissolution of PMMA pellets in the MMA solution took at least 24 hours under constant magnetic stirring. When the components were mixed well, the solution was a light amber color and 0.8 ~ 1 ml of bonding solution was applied on the steel

substrate using a pipette. A PMMA disc placed on the steel substrate was pushed down by hand

to remove bubbles and held on under a pneumatic press for more than 12 hours with 20 psi.

Following the bonding of the PMMA to the substrate (Figure A. 4 (b)), the PMMA disk

was cut down to the desired thickness using a fly cutting machine (Precitech Optimum 120,

Keene, NH) at CAMD (Figure A. 4 (c)). Since a 220µm recess depth in the PMMA mold was

desired, 2,780 µm out of 3,000 µm was fly cut with feed rate and spindle speed setting of 70 (no

unit on the machine). About 200 µm of PMMA was removed in each cycle. Figure A. 5 shows

the fly cutting process for a PMMA disk bonded on the stainless steel substrate.

144

Figure A. 5 Fly cutting process.

A.4 X-Ray Exposure and Development

A 1.3 GeV X-ray lithography beamline for microfabrication (XRLM1) from synchrotron at the Center for Advanced Microstructures and Devices (CAMD) was used as a light source for the LIGA process. In order to deliver sufficient energy onto the bottom of the 220 µm PMMA resist, 4,000 J/cm3 was set as a bottom dose. In the case of X-ray exposure with the graphite mask, less than ratio 5 between the top and bottom dose of the resist was preferred to avoid the resist foaming during the X-ray exposure. 8,629 J/cm3 of top dose on the PMMA resist without additional filter was set with 2.16 top-bottom dose ratio. With the polyimide x-ray mask, a filter

(F-02: 6.8 um Aluminum) was required to maintain the top-bottom dose ratio at less than 5 with a 13,340 J/cm3 (ratio: 3.33) of top dose. Additionally, the contrast ratio between the bottom dose of resist in the exposed area and top dose of the resist under the absorber was desired to be maintain larger than 100. Graphite and polyimide X-ray masks with 11µm of gold absorber under the same exposure conditions given above had contrast ratios of 200 and 333, respectively.

Figure A. 6 (a) shows the DEX02 X-ray scanner (Jenoptik GmbH, Jena, Germany) at CAMD.

Figure A. 6 (b) shows the loaded X-ray mask and PMMA resist on the scanner and clamped to each other with a gap of less than 15µm (Figure A. 6 (c)) for the x-ray exposure.

145 X-ray (a) (b) (c)

Figure A. 6 DEX02 X-ray scanner (Jenoptik GmbH, Jena, Germany) equipped with XRLM1 beamline in CAMD.

The exposed area of the positive PMMA resist was developed in GG developer and rinse solution which were mixture of the chemicals listed in Table A. 3. The 220 µm thick resist was developed through 3 cycles consisting of 20 minutes in GG developer and 40 minutes in rinse solution per cycle with moderate agitation by magnetic stirrer. Following three cycles of development, the PMMA resist was rinsed again by immersion in the de-ionized water bath for more than 12 hours with agitation and dried by gentle stream of nitrogen gas.

Table A. 3 PMMA developing and rinse solutions.

PMMA Volume PMMA Volume

Developing solution percentage Rinse Solution percentage

2-(2-Butoxyethoxy) 2-(2-Butoxyethoxy) 60 % 80 % ethanol ethanol

Morpholine 20 % De-ionized water 20 %

2-aminoethanol 5 %

De-ionized water 15 %

Total 100 % Total 100 %

146 Figure A. 7 shows SEM images of developed PMMA exposed by X-ray through graphite

and polyimide X-ray masks. Depending on the physical properties of mask substrate, the surface

profile of the sidewalls in the developed resist had different characteristics. Graphite is a cost-

effective material [136] as a X-ray mask but its high porosity produced mouse-bite like defects

along the edge of deposited gold absorber and those defects resulted in striations on sidewall of

the developed resist [137]. Taut polyimide membrane provided smooth surface flatness and

(a) (b)

Figure A. 7 SEM pictures of developed PMMA. (a) Developed 100 µm curved channel exposed by X-ray through graphite X-ray mask. (b) Developed 50 µm cross- shaped channel exposed by X-ray through graphite X-ray mask. (c) Developed 100 µm curved channel exposed by X-ray through a Kapton® X-ray mask. (d) Developed 50 µm cross-shaped channel exposed by X-ray through a Kapton® X- ray mask.

147

Figure A. 8 Developed PMMA master for the Ni electroplating (100µm width channel).

produced no mouse-bite like defects. While Figure A. 7 (a) and (b) shows striation on the

sidewall along the developed recess of PMMA exposed through graphite x-ray mask, Figure A. 7

PMMA exposed through the polyimide X-ray mask. Figure A. 8 shows a developed PMMA master readied for nickel electroplating. Before electroplating, a formaldehyde free wetting agent

(E-Liminate Pit, DisChem Inc., Ridgway, PA) [138] was flushed into the developed recess to give good wettability between the electrolyte and bottom of the recess where the nickel was to be deposited.

A.5 Nickel Electroplating and Post Processing of the Mold Insert

E-Form Sulfamate Nickel Concentrate (DisChem Inc. Ridgway, PA) with a 180 g/L nickel metal concentration was used as a nickel sulfamate electrolyte to deposit nickel structures in the recesses of the developed PMMA. E-Form sulfamate nickel concentrate is designed to reduce the internal stress of deposited nickel and meet purity requirements which are critical for the LIGA process [139]. Solution make-up process steps for the new electroplating bath are

148 explained in the data sheet for the E-Form Sulfamate Nickel Concentrate [139]. Nickel metal

concentration was maintained in the range of 90 ~ 120 g/L and boric acid (H3BO4) (DisChem Inc.

Ridgway, PA) was added in the electroplating bath to buffer pH changes during electroplating

with the point of saturation by temperature, 45g/L of Boric Acid for 50 °C bath temperature

[140]. E-Liminate Pit (DisChem Inc. Ridgway, PA) functions as a wetting agent to prevent pitting of the electroformed nickel by hydrogen gas generated during the electroplating [138]. A pH level in the range of 3.8 ~ 4.2 was maintained by adding E-Line pH Sulfamic Acid (H3NO3S)

(DisChem Inc. Ridgway, PA) [141]. Typical compositions of the nickel sulfamate eletctoplating bath used in this work are listed in Table A. 4.

Polypropylene plates were used to build the nickel electroplating bath (4B Plastics, Inc.,

Baton Rouge, LA) with a capacity of 70 liters shown in Figure A. 9. The plating bath consisted of one large and two small separated cells. Apparatus for heating, filtering and pumping for the

Table A. 4 Compositions of nickel sulfamate electroplating bath.

Ni sulfamate bath componets Concentration

Sulfamate nickel (Ni(NH2SO3)) concentrate (180g/L) 89 g/L

Boric acid – H3BO4 (99.8 %) 45 g/L

Formaldehyde free wetting agent 0.3 %

Control pH by Sulfamic acid - H3NO3S (99.9 %) pH: 3.6 ~ 4.0

149 In-tank pump In-tank heater In-tank filter

Temperature Electroplating controller bath

Power supplier

Figure A. 9 Nickel electroplating setup.

circulation of solution were arrayed in the large cell. The two small cells were designed to accept

a plating jig and an anode bag for nickel electroplating. A temperature controller (2110-R3000,

Chromalox Inc., Pittsburgh, PA) with a fluoropolymer covered immersion heater and sensor

(GTF-11, Chromalox Inc., Pittsburgh, PA) maintained the plating solution constant temperature

of at a 50 °C. An in-tank filter system with a self-priming pump continuously filtered solution

through 1 µm and 10 µm wound polypropylene particle removal filter cartridges (Technic Inc.,

Cranston, RI). A chlorinated polyvinyl chloride (CPVC) in-tank pump with a fan cooled motor

(Technic Inc., Cranston, RI) circulated and agitated the solution among cells.

An electroplating jig for the 4.75″ stainless steel substrate made of polysulfone plate

(McMaster-Carr, Atlanta, GA) is shown in Figure A. 10 (a). Nickel pellets (Belmont Metals Inc.,

Brooklyn, NY) for the anode were packed in a titanium basket covered by a polypropylene or

cotton anode bag (Technic Inc., Cranston, RI). The plating jig and anode bag were the working

electrode and counter electrode, respectively, and constant current was supplied to the plating

150 Anode bag

Cathode

(a) (b)

Figure A. 10 Nickel electroplating. (a) Jig for 4.75″ stainless steel substrate with 3.75″ opening; (b) Bath with cathode and anode connections.

cell using a DC power supply (GPR-1810HD, Instek America Corp., Chino, CA). The electroplating cell with the anode basket and plating jig in place is shown in Figure A. 10 (b).

A 10 mA/cm2 current density was used for filling of the micro-meter sized recess of

master and a 40 mA/cm2 current density for overplating for the nickel mold insert. The 0.0146

ampere (A) current for filling the micro recesses with a total plating area of 1.4612 cm2 resulted

in a deposition rate of 10 µm/hour and deposited mushroom-like nickel structures as shown in

Figure A. 11.

Figure A. 11 Mushroom-like nickel forming in the recess.

151 For overplating of the entire resist, it was necessary to make the top surface of the

PMMA resist conductive for the nickel metal to be deposited on it. A 15nm thick layer of gold

(Au) deposited by e-beam evaporation resulted in a 51.5 cm2 area to be electrodeposited. A

constant current of 2.06 amperes (A) produced a deposition rate of around 40 µm/hr, 5 days were needed to deposit sufficient nickel for milling machining of the top of the deposited nickel mold.

Figure A. 12 (a) shows a completed overplated nickel layer on the resist with a coralloid structure along the rim. The overplated nickel variation was mill to 3 mm thick. The overplated substrate was laid in a circular recess on an acrylic plate, and acrylic adhesive (Weld-On 42, IPA

Corp., Compton, CA) was applied and hardened to prevent damage to the microstructure

underneath the overplated nickel during milling. Figure A. 12 (b) shows the flat surface of the

nickel mold after milling. The machined nickel mold was disassembled from acrylic plate by

dissolving the hardened adhesive with immersion in an acetone bath for over 24 hours.

(a) (b)

Figure A. 12 Post processing of the electroplated mold inserts. (a) before, and (b) after milling.

152

(a) (b)

Figure A. 13 Laser welding of (a) mold insert into (b) stainless steel substrate.

(a) (b)

Figure A. 14 SEM images of micromachined brass (a) and (c), and X-ray LIGA (b) and (d) mold inserts.

153 Circular recesses of the same diameter and depth as the nickel mold were milled from

stainless steel substrates (Figure A. 13(a)). The nickel mold inserts were placed in the recess and

flush with the steel substrate. Laser welding (Mezzo Technologies, Baton Rouge, LA) was used

to make a semi-permanent coupling between the nickel mold insert and the stainless steel

substrate. Figure A. 13 (b) shows the final LIGA mold insert, which is compatible with hot

embossing machine at CAMD and CBM2.

Figure A. 14 shows representative SEM images of the nickel microstructures fabricated

by X-ray LIGA process and comparable images of micromachined brass microstructures. The

Brass mold insert structures show curvature at the edge of the microstructure and roughness along the sidewall (root mean square surface roughness of 402.42 nm), while the LIGA microstructures shows sharp edges and smooth sidewall surfaces (root mean square surface roughness of 78.77 nm).

154 Appendix B OPTIMASTM Macro

/////////////////////////////////////////////////////////// // ** Multiphase flows in microchannels ** // ///////////////////////////////////////////////////////////

/* Description: GasBubbleMeasure.Mac – Measurement of gas bubble and liquid plug lengths in a polymer microchannel */

/* Change Log 11/06/07 Modified by Namwon Kim */

/* Program procedure: 1. Load clear field file. 2. Load image file. 3. Clean up the image file dividing by the clear field. 4. Record the information about bubbles to an Excel file. */

/* Most real variables are preceded by an r. Most integer variables are preceded by an i. Most character variables are preceded by a c. */

/* Declare variables and arrays */ integer empty[18]; empty[0..17]=‐1; real Match[20]; cMainDir = GetDirectory( "D:\\Namwon\\optimas\\"); // Assign main directory OpenConfiguration (cMainDir:"\\Config\\2xobj33bin.cfg"); // Load configuration file hChanSheet1 = DDEInitiate ("Excel", "Sheet1"); // Establish link with open Excel file if (hChanSheet1==FALSE) { SHOW("Please check if Excel is open."); Pause(); }

LoadMeasurementSet(cMainDir:"\\Measurement\\microbubble.set"); // Load measurement set BOOLEAN bLoaded = LinkMeasurementSet(2,"sheet1",2,2,FALSE); // Link measurement set to Excel file OpenImage(cMainDir:"\\Frames\\Image\\clrimg\\Clrfld.tif"); // Open clearfield image Calibrate(Xobj33bin); // Switch to calibration set MacroMessage("Please select the channel area."); rchannelROI=SelectROI(); // Returns x, y coordinates of the left, top and right, bottom of the ROI.

XOrigin=rchannelROI[0];

155 YOrigin=rchannelROI[3]; YLeft=rchannelROI[1]; MidChannel=abs((YOrigin‐YLeft)/2); cImageFiles=(FileWildCardList(cMainDir:"\\Frames\\Image\\*.tif")); //Get all of tiff file names in the main directory. iLength=GetOneDimension(cImageFiles,0); Forward=(Prompt("Are the files arranged in a positive counting sequence?",2)); if(Forward) iMarker=0; //Start at the top file. else iMarker=iLength‐1; //Start at bottom file.

/* Write headings to the opened Excel file */

DDEPoke (hChanSheet1, "R1C1", "Frame"); DDEPoke (hChanSheet1, "R1C2", "Area"); DDEPoke (hChanSheet1, "R1C3", "Perimeter"); DDEPoke (hChanSheet1, "R1C4", "Centroid"); DDEPoke (hChanSheet1, "R1C5", "Circularity"); DDEPoke (hChanSheet1, "R1C6", "MajorAxisLength"); DDEPoke (hChanSheet1, "R1C7", "Breadth"); DDEPoke (hChanSheet1, "R1C8", "XYAxisAngle"); r=1; rr=1; /* Loop for forward‐counting *.tif files */ if(Forward) { /* Begin main loop */ while(iMarker

156 hbubblehandles=GetScreenItemHandles(0x0004); //Get a list of screen objects (0x0004: Area) if(GetOneDimension(hbubblehandles,0)>0) { /* Obtain the bounding ROI of each bubble. */ SetExport(ArBoundingROI,1,TRUE); //Set "To DDE" MultipleExtract(hbubblehandles); // Get the data object /* Extract the values of the bounding boxes. */ BoundsY=mArBoundingROI[,,1]; //y‐coordinate BoundsX=mArBoundingROI[,,0]; //x‐coordinate /* Load the rest of the bubble measurement set. */ bLoaded = ActivateMeasurementSet("Bubble Data Set"); if (bloaded==false) beep(); MultipleExtract(hbubblehandles); /* Put information into variables. */ Area=mArArea; Perimeter=mArPerimeter; Centroid=mArCentroid; Circularity=mArCircularity; MajorAxisLength=mArMajorAxisLength; Breadth=mArBreadth; AxisAngle=mArMajorAxisAngle; } /* Determine Bubble Identities */ k=0; //Match vector iteration variable. for(i=0;i

157 r++; DDEPoke (hChanSheet1,"R":TOTEXT(r):"C1:R":TOTEXT(r):"C31", empty); rr++; CloseImage(); //Close Active Image iMarker++; //Increment Counter } //End loop } DDEPoke (hChanSheet1, "R":TOTEXT(r):"C2", ‐4321); CloseImage(); //Close clearfield image. DDETerminate(hChanSheet1); //Terminate DDE

ObjectWildCardList(".*", 2);

158 Vita

Namwon Kim was born in Chuncheon, Kangwon, South Korea. He received his

Bachelor of Science in mechanical engineering from the Kangwon National University,

Chuncheon, South Korea, in 1998. He was with Hyundai Motor Company from 1997 to 2001

as an auto-parts development engineer. He joined at the Department of Mechanical

Engineering in Louisiana State University, Baton Rouge, Louisiana as a graduate student and

currently working toward the doctorate degree with the supervision of Dr. Michael C. Murphy

and Dr. Dimitris E. Nikitopoulos. He is expected to earn the degree of Doctor of Philosophy at

May 15, 2009.

159